Computing Education Things

#45 — An anthropological approach to AI in Computing Education

Daniel Prol — Fri, 22 May 2026 05:02:20 GMT

Reflections

Great conversation at Club Dalroy Podcast between Santiago Huvelle, Juan Serrano, and Francisco Javier Rubio. This topic opens the door to many different kinds of reflection. This discussion approached it more from the perspective of anthropology, the humanities, and the social sciences. A few ideas came up that I hadn’t really heard expressed this way before.

Thread below with some loose notes and reflections from the episode (not literal quotes):

There is no intention, objectivity, spirituality, or semantics inside these models. At their core, they are systems of probability and statistics though operating at extraordinarily sophisticated levels.

Memory becomes irreplaceable at crucial moments. It is a habit that needs to be cultivated, much like public speaking or writing.

The pace of learning matters. Immediacy removes process, rhythm, exposure, depth, effort, and the natural timing of education. It’s almost as if we’ve become scandalized by human limitation itself. One point they raised was that universities seem to produce fewer true intellectual “figures” today: Academics with authority, wisdom, and synthesis because those long, demanding formative processes are disappearing.

One example I found especially interesting: one professor increasingly uses close textual commentary in university classes and requires handwritten notes in secondary education, collecting them afterward. He realized that the effort students must make is so significant that, paradoxically, it ends up generating genuine interest and attachment to the subject even though a machine could technically perform those tasks perfectly.

Recovering the habit of wisdom. To know something through its deepest causes is, in a sense, to love it; and that requires time. Knowing involves participation. Even boredom in the classroom can become meaningful if the will remains engaged because choosing to remain attentive is itself a form of love for knowledge.

The loss of presence. One speaker mentioned that Humanities and Psychology programs at UFV prohibit laptops because they are tired of educational spaces, opportunities for shared presence, eye contact, and genuine dialogue between teacher and student, being interrupted by screens acting as physical and symbolic barriers.

One participant argued that thought is activated in three fundamental ways, all of which AI may interfere with:

Study and reading: reading, rereading, taking notes, pausing, returning to the text, questioning it. A deep process.
Conversation: speaking with students and colleagues. Thinking requires embodiment, presence, and relationship. We lose something essential when materiality and direct contact disappear.
Writing: struggling with the blank page, formulating ideas, expressing them, explaining them aloud. It is through that process of articulation that the mind becomes illuminated.

There is also the risk of mistaking simulation for understanding. AI can create the illusion that you understand something because the output “sounds right,” but the understanding is not actually yours. One speaker described it as “grading the work of nobody.” There is an apparent liberation, but it often requires endless iteration, diminishes the joy of thinking, creates dependence, and can gradually atrophy intellectual habits even if the productivity gains are real. The deeper threat lies in cognitive offloading: thought itself is our capacity to encounter reality as it presents itself.

To educate is an incarnational act. The educator helps illuminate the student’s mind and accompanies the transition from potential knowledge to actual understanding. The professor does this as someone who has already walked that path and can guide students in reproducing it within their own minds. As St. Thomas Aquinas might put it in a Christian key: “In the expressed word there is light.”

To teach or mentor someone is not only to share information, but also to awaken an attraction to the beauty of the subject itself.

A professor is also a witness: someone who opens horizons and brings will, passion, and commitment into the classroom. A true professor becomes an intellectual and moral reference point.

Evaluation methods are also changing: oral exams, interviews instead of traditional papers, more in-class writing, and essays completed on site.

If assessments continue to rely on deliverables that AI can easily solve, the system loses its ability both to measure learning and to truly educate.

Recommendations mentioned during the episode:

If you’re interested in education and formation from a more spiritual perspective, this meditation is definitely worth listening to:

Technology Adoption as a Design Problem

I’ve already recommended Afueradentro here and here before, and I’ll do it again. It’s one of my favorite podcasts in Spanish. Jorge Caraballo is, in my opinion, one of the best interviewers in the Spanish-speaking world right now.

This was a fantastic episode with Rafael Orduz, who has a PhD in Economics from The University of Göttingen (Germany), a remarkable career across both the public sector (senator, deputy minister of education, etc.) and the private sector, and who now dedicates himself to continuous learning and to helping entrepreneurs grow, adapt, and stay ahead through automation workflows and AI.

I’m leaving it here together with some of the ideas that resonated with me the most:

The root of wanting to learn is preserving autonomy.
Sustained reading is becoming harder because of the constant stream of short-form messages competing for our attention.
During his PhD, he became deeply studious because he was entering territories that were genuinely unfamiliar from a learning perspective. He compares that to studying automation workflows today: at first, tools like n8n, Make.com, or Zapier feel like “Serbo-Croatian,” but wrestling with them intellectually becomes enjoyable. Curiosity emerges from both necessity and genuine pleasure in learning.
One fear he had to overcome was teaching highly technical young people. He realized the field is so vast that nobody knows everything. His approach became practical: learn an automation process, build a small portfolio of examples, test them, and apply them professionally. Not knowing something is not a flaw, it simply means others know different things.
He strongly rejects the idea that older adults struggle with technology because of age or biology. In his view, it’s fundamentally a design problem. People need accessible tools, opportunities to practice, and practical learning experiences rather than abstract theory.
Once older adults learn these tools, many experience it almost as a personal revolution: the empowerment that comes from understanding how things work and realizing they can participate in this technological world too.
Heidegger said that “language is the house of Being.” In that sense, interacting effectively with LLMs also requires command of language. Professions centered around language are being profoundly transformed.
There is something deeply valuable about intergenerational learning: older people learning from younger people and vice versa. What matters is creating spaces for conversation. He mentioned the “Intergenerational University” initiative at the University of Caldas as an example.
He also emphasized the importance of connecting initiatives and learning from best practices across regions and institutions: “What are they doing in Manizales that we are not doing here?” There is already a small but meaningful social fabric that could begin to articulate itself more intentionally.
One line I especially liked: sometimes he “steals hours” to read history, even though it may seem completely “unproductive.”
Finally, he reflected on attention and time: he admits he checks his phone too much and has to consciously control it. The challenge is organizing time well, reducing distraction, and overcoming anxiety through action itself by simply doing the work.

Consciousness ≠ increasing computational power

Over the last few days, I’ve been reading very carefully the paper of Alexander Lerchner, Senior Staff Research Scientist at Google DeepMind, on what he calls “The Abstraction Fallacy.”

For example, an artificial heart may appear functionally equivalent because it pumps blood, but a human heart is not exhausted by that single function. It participates in many other physiological processes, releases hormones, and exists within a much richer causal network that the simplified model does not fully reproduce.

In the same way, processing information is not equivalent to generating subjective experience. The mistake lies in confusing the abstract description of a process with the process itself, or in assuming that if we reproduce the functional map of something, we have therefore captured its material reality.

Artificial consciousness remains, at least for now, a somewhat speculative and conceptually confused idea because what we actually have are systems for large-scale data processing: tools capable of reproducing certain high-level cognitive processes such as reading, painting, analyzing, or writing but not entities with genuine judgment or interiority. AI processes symbols, but it is still we who give them meaning and purpose.

This is why the paper also criticizes a certain emergentist faith surrounding AI: the assumption that if we continue adding enough complexity, scale, and computational power, consciousness will somehow emerge at some point. That feels like an unjustified extrapolation. More computation may lead to better outputs, stronger imitation, or more convincing linguistic and cognitive abilities but not necessarily to the emergence of the full human reality itself.

Are you also skeptical about this? I’d be curious to hear your thoughts in the comments.

🔍 Resources for Learning and Teaching Computing

→ A really good free lecture on system design

Watched Vasilis Syrakis’s viral “I Was Laid Off by Atlassian” video on YouTube. Pretty based take.

→ An Introduction to Software Engineering

A great online course pack covering the fundamentals of software engineering for college-level students, created by Prof. Will McBurney and Prof. Mark Sherriff at UVA. Speaking of UVA, I came across their UVA CS Advising Guide and found it really useful.

→ CS51: Introduction to Computer Science (Pomona College)

This course, taught by David Kauchak and Alexandra Papoutsaki, provides an intro to CS through a survey of its major subfields. Topics covered include the history and ethics of computing, computer systems, mathematical foundations, algorithms, theory of computation, programming languages, and machine learning. It assumes basic familiarity with programming while introducing more advanced concepts such as recursion. Programming instruction is primarily conducted in Python.

Other Pomona College courses offered in Spring 2026 that may also be worth visiting include Computational Design Tools and Data Structures and Advanced Programming.

→ Training for Teaching-Focused Careers

Really strong white paper from CRA. It explores how universities can better prepare and support PhD students for academic teaching careers, particularly those interested in teaching-focused roles.

🦄 Quick bytes from the industry

→ CS gives you a highly versatile skill set as the industry continues to evolve

I went on a marathon catching up on SWE Accelerator Podcast interviews with Aman Manazir. Here are my thoughts of this interview:

Being a software engineer is about far more than writing code. It’s about building reliable systems, handling edge cases, and navigating real stakeholder constraints. If you’re a software engineer or CS student right now, stay with it. There’s still huge learning potential in the field. You just have to adapt quickly, especially around AI.
Even with all the uncertainty, a background in software engineering or computer science gives you skills that go beyond coding: critical thinking, abstraction, problem-solving, and systems thinking. That’s an incredibly versatile foundation if the industry keeps changing.

🌎 Computing Education Community

Interesting postdoctoral opportunity at the University of Michigan School of Information with Nazanin Andalibi for scholars working at the intersection of computing and society.
Tarek Salah Uddin Mahmud is recruiting a fully funded PhD student at Texas A&M–Kingsville for Fall 2026 in AI + Software Engineering, cybersecurity, and software quality assurance.
Robert Ross is recruiting a funded PhD student in Human–Robot Interaction.
Yong Suk Lee is hiring a postdoctoral fellow at Notre Dame to study AI’s impact on work, labor markets, and agentic AI as a social science research tool.
Fred Martin shares UT San Antonio’s “AI for Everyone!” middle school summer camp (June 1–4), featuring ML, micro:bit projects, and GenAI ethics.
Ying Cheng announces a postdoctoral opening at Notre Dame focused on learning analytics, psychometrics, and interdisciplinary education research. Deadline June 12.
My WG colleague Maíra Marques has written a series of really thought-provoking LinkedIn posts about the four different “hats” she wore this semester (teaching, undergraduate program director, research, and service). Highly recommend them.
Majeed Kazemitabaar is recruiting students.
CCSC Northwestern 2026 (Oct 9–10, Whitman College, Washington) invites papers, tutorials, panels, workshops, and student posters across computing and CS education topics. Main submissions due June 30, student posters Sept 27. Shared by John Stratton.
AccessComputing is hosting a discussion-based book club on Digital Accessibility Ethics (May 21–June 11), exploring disability inclusion across AI, cybersecurity, healthcare, design, gaming, and broader technology practice. Shared by Brianna Blaser.
ICCE 2026 (Nov 30–Dec 4, Christchurch, New Zealand) is accepting submissions across seven subconferences, with the call for papers deadline on May 31. Shared by Tim Bell.
A pre-tutorial at ISCA 2026 (June 28, Raleigh) will explore practical strategies for recruiting, onboarding, and mentoring undergraduate researchers in computing. Shared by Susan Rodger and Kelly Shaw.
Koli Calling 2026 (Nov 5–8, Finland) invites full papers, discussion papers, posters, and demos, with full paper submissions due July 12 (or July 19 with abstract pre-registration). Shared by Rodrigo Duran.
Akesha Horton is recruiting computing educators for a study on redesigning authentic assessment in the age of GenAI, including interviews and collaborative assignment-design workshops.
AI4CAREER, a workshop on responsible AI for STEM career development in K–16 education, invites short papers and motivation statements on AI, career pathways, learner agency, and equity ahead of its June 27 event in South Korea. Shared by Si Chen.
Virginia Tech’s Echolab is recruiting CS instructors who use live coding for a paid study ($50) evaluating AI-generated lecture handouts derived from coding sessions and their pedagogical usefulness. Shared by Yuhang Zheng.
The Journal of Open Source Education (JOSE) is seeking reviewers to help reduce a backlog of 60+ submissions, including open-source teaching tools and classroom resources across disciplines. Shared by Zachary Kurmas.
ACM TechTalk (June 4) will feature Val Andrei Fajardo on building LLM agents from scratch covering tool abstraction, MCP integration, agent loops, and open standards without relying on black-box frameworks.
Researchers from UNC Charlotte and Embry-Riddle are recruiting instructors for a short survey study examining how student-centered practices are reflected in course syllabi. Shared by Nadia Najjar and Debarati Basu.
EduHPC 2026, co-located with SC26 in Chicago, invites papers and assignments on high-performance and parallel computing education, including AI impacts, workforce development, pedagogy, and classroom practices, with submissions due July 17. Shared by Suzanne Matthews.

🤔 Thought(s) For You to Ponder…

I found this interview with Jorge Carrión, who also has an academic side to him, really thought-provoking.

Here are some of the ideas that stood out to me the most:

1/8 Creativity and originality are becoming increasingly important in contrast to formulaic, easily reproducible content that AI can quickly imitate. In this context, the role of the content curator starts to make a lot more sense. When everyone can generate endless content and ideas, the real value shifts to the people who can curate, filter, and make sense of what already exists.

2/8 Instead of using human-AI collaboration to create beauty, humor, criticism, or high-quality audiovisual and written work, we’ve filled the digital space with “AI slop,” visual junk, and instant-consumption social media content, flooding the market with extremely low-quality material.

3/8 There is also growing human distrust toward AI. The moment we work with it, the sense of lived experience, documentation, and factual grounding seems to disappear. Everything generated by AI is often perceived as fiction. Many AI-generated texts feel detached from reality or lacking in substance — the same way those glossy, artificial-looking images do. That perception has partly prevented us from truly co-creating with these tools.

4/8 All streaming platforms ultimately rely on something deeply human and timeless: at the end of the day, we like having stories told to us.

5/8 We need to stop obsessing over trends and virality. The real work lies in storytelling and in capturing attention meaningfully. Readers are looking for distinction and singularity — not formulas. The challenge is preserving and cultivating that uniqueness while also making it economically sustainable.

6/8 We’re returning to theaters, cinemas, gyms, and other physical spaces because we need shared experiences and meaningful in-person connection.

7/8 Education matters. Study matters. We need to understand something before entering systems capable of reshaping our perception of reality.

8/8 The celebrity magazine Pronto has shut down its website and will now exist only in print because it does not want AI systems reading and profiting from its content. That’s obviously not a viable strategy for every publication — they already have a large print readership, and it doesn’t guarantee survival — but I still found it fascinating. It forces us to rethink strategy in this new landscape and reconsider how to differentiate ourselves.

Definitely worth taking some time to go through this thread.

There’s a mobile psychology clinic for people experiencing homelessness. An inspiring initiative by the cooperative Lotura Giza Garapena. They travel across San Sebastián and nearby towns providing mental health support to people living on the street or without stable housing.

They experience it as a sense of relief.
Sometimes we communicate using Google Translate or real-time translation tools focusing on small steps, little by little.

Reading on paper is important for several reasons:

It allows for slower, deeper reading, with real time and attention. It creates distance from the noise of social media and constant overstimulation.
It helps us follow lines of reasoning, reflect on ideas, and move from thought to thought instead of simply chasing sensations.
Reading also strengthens the imagination: the images are not handed to us.

Speaking of deep cognitive exercises, writing is one too, according to this Psychology Today article by Wendy Lustbader. Every time you sit down to write, your brain has to simultaneously coordinate memory (to retrieve facts and ideas), reasoning (to structure and connect concepts), and meaning-making (to give coherent and clear expression to what you think). That integrated act of thinking helps “wire” the mind toward clearer and more organized forms of thought. Writing is thinking with greater rigor and presence. It is not just about transcribing ideas, but about confronting them, organizing them, evaluating them, and transforming them into something understandable for others and for yourself. The constant practice of translating abstraction into language requires attention and mental engagement, activating neural networks involved in analysis, synthesis, and metacognition. All of this is to say that, in this context, using LLMs as support tools, although extremely useful for many tasks, can encourage a more passive “consumption” of thought if used without balance. When you delegate the generation of text to an AI system without actively participating in the process, you lose the opportunity to train those internal capacities. The machine may provide clarity efficiently, but your brain no longer practices the act of thinking through and organizing ideas itself.

Read this article by Irene Dorta in El País:

Nothing particularly new in the legal world: AI is mostly being viewed through a practical lens, improving efficiency and automating workflows, while the usual concerns about data and potential misuse remain front and center. The general feeling seems to be that a basic level of AI literacy is enough for now.

The same trend is playing out in medicine (American Medical Association). More than 80% of U.S. physicians now use AI in their work, compared to less than half just three years ago. Attitudes are changing quickly as well: over three-quarters of doctors believe AI improves patient care. AI is increasingly valued for reducing the burden of routine tasks, though concerns remain about loss of skills, data privacy, and the potential impact on the physician-patient relationship.

I love election nights. I really enjoyed following the coverage of the regional election night in Andalusia, one of Spain’s largest autonomous communities, with Carlos Alsina and his panel on Onda Cero’s YouTube channel. I’ll definitely be following the 2027 Spanish general election with them as well. I also thought Ignacio Varela made a very good point: in polling, what really matters is vote share and percentage of the vote, not seats, since seat projections are ultimately an artificial modeling assumption. The polls were actually quite accurate within the margin of error.

Excellent summary by George Jacobson about Pedro Ballester, who is on the path to sainthood. Really enjoyed it. If you’re interested in his life, I’d highly recommend a short book about Pedro called I’ve Never Been Happier. It’s excellent.

Send the arXiv AI-generated slop, get a year long vacation from submissions. Jay Peters (The Verge).

I’m a huge fan of the comparsas from the Cádiz Carnival. This was a really interesting glimpse into the creative process of Manuel Cornejo. Thanks to Daniel Fopiani for bringing this to my attention. Two things he said really stuck with me: he feels a responsibility to give something beautiful to his audience through the music, lyrics, and overall idea and because they never know how their performance will land each year, they’ve learned to enjoy the journey itself. Honestly, a great philosophy.

Manuel is also a high school teacher, and I really appreciated his optimistic perspective on education. The host mentioned that his sister is a university professor and feels like everything is changing dramatically because of AI, almost with a sense of frustration, as if saying: “A few years ago I really enjoyed my job, and now, with AI, it feels like your role is mostly just guiding students.” Students rarely ask questions anymore; whenever they have a doubt, they immediately look it up on a computer. But Manuel says he genuinely enjoys his students and his relationship with teenagers that they keep him feeling a little younger. He likes stepping into their world and finds a lot of joy in that. He also says he loves teaching and sharing knowledge, and that he believes education is something beautiful. Of course it changes, but everything changes. As a teacher, he says, you have to adapt to the times your students are living through and help them learn how to navigate the world well to become capable people with critical thinking skills. The goal is to do what’s best for them, both academically and personally: to make them think and encourage them to see things from different perspectives. He also points out that part of becoming a mature teacher is understanding the context in which a child is growing up, the conditions at home, the challenges they may be facing and trying to help however you can.

It’s official. Can’t wait to dive into it.

La Casa de Carlota in Barcelona is doing inspiring work in inclusion and employment opportunities for people with Down syndrome, autism, and other intellectual disabilities. Talent has always been viewed through a very ableist lens,” they point out. Their approach is to focus on people with talent because not everyone is the right fit and they also actively recruit and select candidates. I like that they don’t take the easy route. They set high standards. They know that combining diverse cognitive abilities and channeling them into creative work is what makes them unique, but they’re also clear that not everything goes and the final product still has to be strong.

Interview by María de la Peña Fernández-Nespral with Paloma O’Shea in XL Semanal:

Some ideas that stood out to me:

Music helps you live better.
I would love for the new building to include a cafeteria where students and professors could have breakfast and lunch together. That way, life continues beyond the classroom: conversations about the world, about music. Menuhin used to speak about the master-disciple relationship rather than simply professor-student.
I take great care in selecting professors; being with them is what I enjoy most. People say I have a good eye for choosing the right individuals. Teachers must be among the very best at their instruments, but they should also be good people, intelligent, and humble. If someone is arrogant, something is not working.

In the same issue of XL Semanal, Carlos Manuel Sánchez interviews José María Jorquera, a CS PhD from the University of Murcia working on security and trust at CyberDataLab:

The weakest link is still the human being.
Almost everything that is protected today will eventually be decryptable once quantum computing matures.
Some technologies are already combining classical and post-quantum cryptography to create a kind of dual-key protection. Satellites have already made that leap. But 100% security does not exist. We have to learn to live with uncertainty.

Are we entering the acceptance phase now?

Andrej Karpathy has joined Anthropic. One of the most influential researchers in deep learning, a founding researcher at OpenAI, and former Director of AI at Tesla. He’s easily one of the most respected voices in AI today. The team Anthropic is putting together is honestly amazing.

Last week I already opened the newsletter with Prof. David Malan, so I don’t want to repeat myself too much, but he’s always a joy to listen to. This episode of the Harvard MCS Career 411 Podcast is also highly recommended. I really connected with what he calls “educational theatricality” or teaching stylistically, a skill he traces back to his admiration for theater people and the kind of experience they create for an audience:

When it comes to my teaching style, I’m not sure how much of an impact it had, because I wasn’t exactly a theater kid in middle or high school, but I did spend a year or two in drama club. And I really loved it: the people, the atmosphere, the whole experience theater people create for an audience. Looking back, I think that may be part of why I never understood why education couldn’t feel more like that. Not entertainment exactly, but something engaging, inspiring, stimulating, a shared experience among everyone in the room. I think it’s those kinds of moments and unexpected experiences in completely different fields that ended up shaping what I do now.

New Fire (Maria C. Escobar): Whether you’re newly married, in the middle of married life, or many years into it, you can find some great resources here for growing in marital love.

Interesting point from Stephanie Palazzolo (The Information) about the expert data that major AI companies are already signing agreements to acquire in order to train AI agents capable of performing expert-level tasks.

Puerto Rico’s political and cultural struggles as expressed through its music. This episode of the No es el fin del mundo podcast by the El Orden Mundial team does a great job exploring it:

Here’s the playlist in case anyone’s curious.

Interesting interview with Santiago Schnell, now Provost of Dartmouth. Highly recommend this other article by Schnell as well. He starts from the idea that education is about forming someone capable of thinking, judging, and taking responsibility for what they affirm. No machine can replace the acts that shape a mind: paying attention, comparing evidence, sustaining a question, defending a thesis orally, recognizing what one does not know, or taking responsibility for the truth. AI should elevate that mind.

I came across this Pew Research Center study and found it really interesting to see how teenagers think about AI. After all, they’re the future of universities. Honestly, I’m glad that many of the concerns about AI’s negative impact revolve around overreliance and loss of critical thinking. If students themselves are already aware of those risks, maybe that will naturally push them toward a more balanced relationship with these tools.

There was a period in my life when I worked in what used to be called growth hacking, then growth marketing, outbound, product marketing… I’ve never been entirely sure where one ends and the next begins. What’s interesting now is that ChatGPT is gradually evolving from being just a product into something closer to a full advertising platform as well. Almost like a traditional Ads Manager, with budgets, bidding systems, banners, campaign management, and performance metrics. Ads will only appear on certain plans, but still, it’s a familiar trajectory. Most platforms begin with a cleaner, less saturated interface, and eventually drift toward the advertising incentives that shaped search engines and social media.

📌 Research Corner

When I explain my research, I always say that my goal is not to put more screens in the classroom, because that does not necessarily lead to better learning outcomes and may even weaken essential cognitive processes such as effort and concentration. What interests me instead is developing more visual educational tools that support computing educators and create better learning experiences for computing students. The integration of these tools does not necessarily have to happen directly inside the classroom, it can also take place in online or remote contexts. In many ways, as this Fortune article by Sasha Rogelberg suggests, the challenge is less technological than it is about integration.

Speaking about my research, this week has been less about implementation, that part will come in two weeks at Aalto, and more about thinking deeply about the problem space and the motivation behind my current project. I touched on this earlier in today’s second reflection. The design problems created by AI are present in academia as well. AI-generated work is increasingly impossible to detect, and fighting against the technology itself does not seem meaningful to me. One alternative, or at least part of the solution, is to redesign the way we assess learning. I think we need to move back toward forms of evaluation where the process matters: real-time thinking, oral defenses, in-class writing, and more authentic assessment practices. Not to avoid AI, but to ensure that learning is still actually happening. Because AI does not eliminate the need to learn but it does eliminate many of the traditional ways we used to demonstrate learning. And that is where universities risk losing their relevance. My tool is specifically designed to capture events such as attempts, block rearrangements, and moments of struggle, because those are traces of real thinking happening in real time. What interests me is not the final grade itself, but how the student arrived there.

Later today I have a catch-up meeting with my ITiCSE Working Group (btw, see you in Madrid at the conference in July!).

This week I’ll also be in full paper-reviewing mode for the assigned SIGCSE Virtual 2026 submissions.

Two really good discoveries this week: Learning in Context and Earli.

🪁 Leisure Line

I’ve been spending the last two weeks working from my parents’ hometown. It’s a place that really brings out my obsession with memories. I visited a few spots that transported me back to moments from my childhood:

On Saturday, we went with my parents, my brother, and his girlfriend to a Galician furancho, a family-run tavern where people sell homemade wine and simple traditional food. The food was amazing. The place also has an old muiño, one of the traditional watermills you find along rivers in Galicia.

On Sunday, I had breakfast with my mom with beautiful views from the café at the Louxo Hotel on La Toja Island.

On Tuesday, I got to spend some time with my friend Marco, who’s building software from O Grove for the world. Always great catching up with him. He’s currently looking for a new role, and I’m sure he’s going to crush his next adventure whatever it ends up being.

📖📺🍿 Currently Reading, Watching, Listening

Saw Pixar’s Hoppers this week. Reinventing the classic environmental “humans vs. animals” story is tough, but I think the movie manages to do it without coming across as preachy because it’s more interested in opening a conversation than delivering a lecture. It’s packed with twists, moves at a great pace, and has a lot of humor and heart. Definitely a fun family movie. I also noticed some evolution in the animation style, which I really appreciated, while still preserving the core Pixar feel. It’s not a masterpiece or anything revolutionary, but to me it does recapture a bit of the magic of old-school Pixar. I really hope they keep moving in this direction.

F1 is back in Canada this weekend. Hopefully we get another race like Miami, great on-track action, real racing, and a sign that the worst of the early races is behind us. Canada has almost everything: high-speed sections, heavy braking zones, demanding chicanes, tons of speed, and that constant feeling that you’re one mistake away from the wall. It has that sense of risk that makes F1 special. Personally, I’m getting a bit tired of all the discussion around the “new F1,” the regulations, and the technical side of things. Since Miami, I’ve been hoping the focus shifts back toward the human element, the drivers, the battles on track, and the real emotion of Formula 1. For example, I’m really looking forward to the Antonelli vs. Russell battle, seeing whether Mercedes takes a real step forward with the latest upgrade package, or whether Ferrari and McLaren stay close enough that the field tightens up. I’m also curious to see if Carlos can keep the momentum going and whether Aston Martin can finally climb out of the hole they’ve been stuck in.

This interview with Erik Varden left me thinking long after I finished reading it. I found myself going back to several of his answers. He always seems to offer a fresh perspective. Sharing it here in case it resonates with you as deeply as it did with me.

💬 Quotable

A human life is a drawn-out affair. And things take time. Great things take time. That’s a principle that Newman liked to stress. And to be human is a great thing.

― Erik Varden

🌐 Cool things from around the internet

A collection of links to stuff I think are worth sharing.

🔗 Bose SoundLink Micro Portable Speaker (2nd Gen) — it probably won’t surprise anyone that one of my hobbies is listening to podcasts and learning from them. While looking for a speaker for a car I drive in Houston that doesn’t have Bluetooth, I came across the Bose SoundLink Micro (2nd gen), and honestly, the sound quality is amazing. Definitely going to get a lot of use out of it.

🔗 Sketchplanations — if you’re into cartoons, take a look at this artist (and a great talk here, about his process). He explains ideas through drawings in a really engaging way.

🔗 Cristine Bedfor — absolutely love this guest house in Seville.

🔗 Fellow — kind of obsessed with Fellow products right now.

🔗 It Builds in Seconds — entire buildings going up in seconds. Fascinating account.

Issue #45 of Computing Education Things was written while listening to:

🔗 Quick Links

🎧 Listen to Computing Education Things Podcast

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

If you’re enjoying Computing Education Things, please like, comment, or share this post! You can also support this work through Buy Me A Coffee. And if you’re finding value in the newsletter, consider forwarding it to a friend, colleague, or family member to help the community keep growing. Thanks for reading!

Buy Me A Coffee

#44 — How to keep people engaged for hours

Daniel Prol — Fri, 15 May 2026 12:19:21 GMT

Reflections

David J. Malan is a Harvard professor best known for taking the online CS50 course to another level. Ryan Peterman interviewed him, and they talked about CS50, relevant skills, engaging classes, and the impact of AI on computing education:

I’ll focus on three of the topics they discuss in the conversation:

How to Make Long-Form Lectures Engaging (Online and In-Person)

From a pedagogy standpoint, CS50 is a three-hour class that people watch online, which means students have absolutely no obligation to stay. And yet they stay, and more importantly, they remember it. But this doesn’t only apply to remote learning, it works in person too. Why?

Malan attributes it to what he calls memorable moments. The phone book example illustrates this well. He tears a phone directory in half to demonstrate the efficiency of binary search. Almost nobody today uses a physical phone book anymore, and yet the example still works because it translates an abstract algorithm into a physical gesture. The irrelevance of the object itself doesn’t matter because the structure is analogous, and that is enough.

He says the goal is not entertainment for entertainment’s sake, nor is it simply about engagement or keeping people’s attention, although there is definitely a performative element to avoid boredom. The deeper goal is to create cognitive anchors: things students can latch onto in their memory.

The brain retains information better when it is attached to a concrete image, a physical event, something that breaks the monotony of abstraction. When a computing student sees an older student physically acting out the bubble sort algorithm on stage, they suddenly have somewhere to place the concept. Later, when they forget the formal definition, they may still remember the image. And from that image, they can reconstruct the definition.

What Malan is really doing is teaching through metaphor. These structures allow students to understand something new in terms of something already familiar, such as using physical activities to demonstrate sorting algorithms and help students build intuition for how those algorithms work. Each of these gestures communicates the same underlying message: “This concept that seems arcane is actually something you already understand, just under a different name.”

They also discuss the tradeoffs of this method and the tension between theatricality and content density. If you spend time making binary search memorable, perhaps you are sacrificing the amount of material you could otherwise cover. Students with prior experience notice this sometimes, and some even express frustration about it. Malan’s position is that the tradeoff is worth it because the goal is not to maximize the number of concepts transferred, but to maximize the number of concepts that actually stick.

A dense lecture with no interaction that nobody remembers has a worse pedagogical ROI than a more memorable lecture that creates durable cognitive anchors.

Finally, Malan admits that much of his energy on stage comes from not wanting to stand in front of a bored audience. He genuinely cares about what happens in the room. Motivation itself becomes part of the content. The hours students will spend later that week working on problem sets require fuel. That means the lecture is not just a moment of instruction, it is also a moment of emotional contagion. If students leave a three-hour lecture without wanting to explore the material on their own, then the lecture failed, even if they technically understood everything that was explained. That is the real goal: for students to walk back to their dorms excited to work on their problem sets. That enthusiasm, that sense of passion is not an accident or simply a personality trait. It is the result of a class intentionally designed to produce it.

The knowledge that will matter most in the future

What’s the point of learning C or pointers if you’re never going to use them in production? For Malan, the answer is epistemological. The distinction he draws is between using and understanding. A software engineer who does not understand the layers of abstraction they work with is not a complete engineer, because they are operating tools without understanding the machinery underneath.

What Malan is really arguing is that the knowledge that endures is not knowing how to implement a hash table in C, but having built one yourself and understanding why it exists, what problems it solves, and what tradeoffs it involves. That experience of building something from scratch, even if you never do it again, shapes an intuition that for example helps you understand higher-level languages later on. It is the difference between someone who knows that dict exists in Python and someone who understands what lies beneath that single line of code.

What CS50 is trying to produce is not merely programmers, but engineers and citizens capable of reasoning from first principles. The language itself is always an implementation detail. What matters is the ability to decompose problems, diagnose symptoms from the foundations upward, and design solutions with an awareness of system constraints.

This distinction becomes especially urgent now that AI can generate code with a level of competence that already surpasses most junior developers. AI can write the code. What it cannot do is decide what should be built, why it should exist, which architecture makes sense, or what kind of user experience matters. System design, choosing the right data structures, deciding which data is worth collecting based on future business problems that remains deeply human territory.

Malan also has historical perspective. He lived through the dot-com crash, the blockchain boom, the hype around AR/VR and Google Glass. Each time, the tech world reacted dramatically. And each time, the field eventually stabilized, with no shortage of new problems to solve. The world only becomes more technological, not less. What changes is the nature of human work within that world.

How AI Has Changed Computing Education

David Malan teaches an introductory computer science course at the exact moment when AI tools can complete problem sets better than most of his students. And he has been thinking about this since before ChatGPT became mainstream. GitHub Copilot came first with autocomplete.

His response operates on several layers, and they are worth separating.

The first layer is creating a differentiated environment. CS50 has its own AI assistant which is intentionally designed to be less useful than ChatGPT. It does not solve the problem for students; it guides them toward the solution. The line is clear: students can use cs50.ai as much as they want, but the moment they open ChatGPT, they knowingly cross a boundary. Instead of trying to detect cheating after the fact, the course redesigns the environment so that cheating requires a deliberate decision rather than simply following the path of least resistance.

The second layer is honesty about the limits of detection. Malan acknowledges that they are not necessarily seeing more cases of academic dishonesty, but what has changed is the ability to prove it. Previously, instructors could point to a URL and say: “this code was copied from here.” AI-generated code no longer comes from a traceable source, it comes from the synthesis of everything. What remains are indirect signals: inconsistencies with a student’s previous work, unusually sophisticated implementations, or perhaps the most revealing detail of all: the code solves last year’s problem set instead of this year’s.

The third layer is the most important one long term: what does it mean to learn programming when AI can already program? Malan mentions that the very same morning he had been prototyping something with Claude. The collaboration was useful precisely because he already knew enough to recognize that the generated solution was 90% correct and to have the technical conversation necessary to fix the remaining 10%. That remaining 10% is exactly where first-principles knowledge becomes indispensable. Without it, “mostly correct” code becomes production software containing subtle bugs nobody understands.

In Malan’s view, AI automates the least interesting parts of the job: writing unit tests, reading unfamiliar API documentation, implementing boilerplate. What it does not replace is design: deciding what to build, for whom, under which constraints, and with which system architecture in mind for future needs.

Even if software implementation itself became fully automated, problem solving and systems thinking would still matter. Those are precisely the cognitive skills that allow humans to collaborate effectively with AI rather than merely consume its outputs.

There is also a broader psychological effect on enrollment. Fewer students are entering computer science, and the reasons are worth distinguishing. The fear that AI will devalue their skills is the obvious headline, but Malan points out that the contraction of the job market specifically the disappearance of junior engineering roles preceded the AI wave and is independent of it. Companies simply stopped recruiting entry-level engineers before AI became the dominant conversation. When the AI wave arrived, it confirmed fears that were already forming. The two causes are related but not the same, and collapsing them into one obscures what is actually happening in the market.

Malan’s response is that this is simply another cycle in a history he has already seen repeat itself: the dot-com crash, blockchain hype, AR/VR enthusiasm. Each time, the field absorbed the disruption and continued expanding because the world keeps becoming more technological, not less.

Taste vs. Criterion/Judgement: The Definitive Take

After months of thinking about the difference between criterion/judgement and taste, this post by José Luis Antúnez feels like the missing piece that finally made it click for me.

With his permission, I’m just going to translate it literally into English, because I think it’s brilliant and there’s really no better way to say it:

For someone without formal training in a discipline, taste is autobiographical. It’s a passive accumulation of cravings shaped by where we grew up, what we’ve consumed, and what happens to be trendy. Challenging an “I like it” risks making someone you care about feel as though their identity is being rejected.
Judgment, on the other hand, is historiographical. It’s an active accumulation of reasons that can eventually produce an immediate response we call instinct. It’s knowledge trained over years, nourished by principles, successes, mistakes, and even biases (no one escapes them), all of which shape a point of view that can be argued according to needs, contexts, budgets, and constraints.
If you lead teams, avoid saying “I don’t like it” without explanation. Professional judgment doesn’t eliminate subjectivity entirely, but it does discipline it. We should learn to reason through what we feel without falling into self-deception.

The “Flaky Compiler” Metaphor

Bryan Cantrill and Adam Leventhal invited Brown University professors Kathi Fisler and Shriram Krishnamurthi onto Oxide and Friends to discuss AI in computer science education. The conversation focused on an experimental spring course built around agentic programming, designed to expose students to LLM-powered “flaky compiler” workflows. The goal was to let students experience failures firsthand, develop testing and type-based safeguards, and help faculty rethink introductory CS curricula.

Their motivation was that tools like Claude have already made it impossible to teach computer science the same way it was taught before. Rather than resist that reality, they wanted to collect real data on how students actually use these systems and experiment with new teaching approaches ahead of the Fall 2026 semester. They also pointed to a major “expert blind spot”: instructors often have very little visibility into what students already understand or misunderstand about LLMs.

To explore this, they launched a small pilot course with 20 students selected from 80 applicants. Participants were required to have at least one semester of programming experience. The idea was not to teach students how to code from scratch, but to investigate how far they could go when working with programming agents as collaborators.

Shriram described agentic systems as a kind of “flaky compiler”: incredibly powerful, but fundamentally non-deterministic. The course was therefore designed to teach students how to get reliable results from unreliable systems by leaning heavily on core software engineering principles. Students learned to write precise specifications, design robust automated tests, use types to make assumptions and interfaces explicit, and rely on proper libraries and constraint solvers instead of naive AI-generated solutions. Verification, code review, iterative refinement, and engineering discipline became core practices.

“Flakiness” itself is not new. Software engineers already deal with flaky systems, flaky coworkers, and even flaky versions of themselves. Agentic systems simply introduce another source of unpredictability that has to be managed responsibly. The class embraced this philosophy by operating almost like a design studio: iterative, experimental, and intentionally messy “like assembling a plane mid-flight.” Students were encouraged to use agentic coding tools, run into failures, reflect on them, and then learn the engineering concepts needed to address those failures.

The instructors shared several assignments from the experiment:

Tetris sequence: Students first used Claude to generate a working Tetris implementation in a single shot. However, when they were asked to build variants such as reversing gravity or dynamically toggling it the limitations of the generated solutions quickly became obvious. This pushed students to think more deeply about representation, architecture, testing, and design.
Data analysis assignment: Students started with a small CSV dataset and later scaled up to much larger datasets. The generated solutions often failed to scale properly, creating opportunities to teach databases, performance considerations, and testing strategies.
Requirements checker project: Students built a checker for Brown CS degree requirements, exposing ambiguities in the specifications and highlighting the complexity of grading infrastructure and formalized requirements.

The teaching approach consistently emphasized types as a constraining and reviewable artifact. Instructors encouraged “types-first” workflows in TypeScript so that prompts and generated code would remain understandable, constrained, and easier to review.

A particularly interesting aspect of the course was its emphasis on peer critique, borrowing from the “crit” model commonly used in design disciplines.

Assessment relied heavily on testing plans, design documents, peer critiques, code reviews, and reflection journals written by the students themselves rather than generated by LLMs.

Students initially reacted with amazement at what LLMs could produce especially after seeing one-shot Tetris implementations but that excitement quickly gave way to a deeper awareness of the systems’ unreliability once the requirements became more complex. Many students reported feeling conflicted about relying too heavily on LLMs for learning, and some noticed that excessive dependence reduced their own engagement and understanding. Usage patterns varied widely: some students leaned heavily on the tools, while others deliberately avoided them on principle.

One surprising outcome involved API usage and costs. The instructors had previously purchased a substantial amount of API credits for another course around $2,500 worth but actual student usage ended up being far lower than expected. This challenged the assumption that students would simply outsource everything to the model whenever possible.

The key thing is that students are like, “Look, I’m taking this course to learn something. If I outsource everything to GPT, I’m not really going to learn the thing I came here to learn.

The discussion also touched on faculty concerns about cheating. The instructors acknowledged the issue but argued that students who rely too heavily on AI ultimately undermine their own education. In practice, many students still wanted to genuinely learn and were hesitant to fully hand over the learning process to AI systems.

It’s also about the job market and everything going on right now. Students are really worried that they need to look good on paper or they’re not going to get hired. So there’s a lot of anxiety around grades. And if you know some of your classmates are using LLMs to get assignments done while you’re doing them on your own, you start wondering: if that lowers your grade, what are you supposed to do for your career? That’s definitely something we hear from students too.

Another major debate centered on when and how agentic tools should be introduced into the CS curriculum. Some students argued that novices should not have access to these tools early on, while the instructors explored several instructional models, ranging from fully agentic courses to hybrid approaches where AI use is more constrained. At the same time, they saw significant upside: because routine coding tasks can now be generated quickly, instructors may be able to introduce higher-level software engineering concepts such as testing, usable security, distributed systems, and multi-user design much earlier in the curriculum.

To me, the whole point of undergraduate education is, in some sense, to break you down and rebuild you. – Bryan Cantrill

The conversation also explored broader implications beyond CS majors. Students who take only a single programming course may become dangerously overconfident if they are taught to trust LLM-generated output without learning the principles of verification and testing. The instructors worried that if computer science departments fail to teach rigorous, and trustworthy AI-assisted development practices, that gap could instead be filled by youtubers lacking sufficient engineering rigor.

We need to communicate to students what learning actually means.

Operationally, the pilot course intentionally kept enrollment small and relied on highly interactive teaching methods rather than heavy automation. Students participated in peer reviews, kept reflection journals, developed testing plans, completed live 20-minute code review finals, and engaged in frequent day-to-day interactions with instructors.

By the end of the course, many students reported a deeper appreciation for software design, testing, and the importance of questioning agent outputs rather than trusting them blindly. Some even became more cautious about using LLMs after the experience. The instructors plan to publish their findings, redesign future courses, and launch additional pilot classes for beginners to better understand how agentic tools can be introduced safely and effectively.

The hosts closed the discussion by praising Brown’s willingness to experiment, its institutional flexibility, and the trust it places in students. The overall tone was optimistic: computer science education is changing rapidly, and universities have a responsibility not only to acknowledge that shift but also to teach students how to use these systems critically, rigorously, and responsibly.

What happens if the software survives, but no one remembers why it survives?

I really appreciate Camilo’s clarity in both his explanations and his choice of words. You can clearly sense his academic and well-read background.

This episode explores a concept that was new to me, but incredibly thought-provoking: epistemic drift, the gradual degradation of the software’s underlying justification. AI makes this problem even more acute because it’s becoming harder and harder to capture those traces as we keep adding features with agents. Those paths of reasoning get lost, and that’s exactly the “justification” Camilo is talking about.

Watch the episode to hear him explain it in depth. Here’s the short version: imagine six features that all still preserve their functionality, but the reasoning behind why something works or how we arrived at that system in the first place can no longer be reconstructed.

The danger is ending up forced into a rewrite because the code still works, yet we no longer understand why or how. And this goes beyond technical debt or code aging: epistemic drift can happen even when everything appears to be working correctly.

Camilo also ends on a more optimistic note about agents: they can help capture traceability (inputs, outputs, steps, and interactions) and give us better observability tools.

To me, this is a wake-up call for AI companies: an agent shouldn’t just satisfy requirements, it should also preserve the justification behind the system.

Could we lose control over the software we’re building?

🔍 Resources for Learning CS

→ Some thoughts on working with AI models

Some Claude content from Eugene Yan. Really good. Thanks for putting it out there. Learned a few things: context as infra, taste as config, verification for autonomy, scaling via delegation, and closing the loop.

🔍 Resources for Teaching CS

→ In-class writing instead of a semester-long paper

Here is a thoughtful episode about writing that focuses on its role in supporting thinking rather than simply producing an artifact. It explores how that role should shape the design of student writing, with more of the work happening in class as a supported, iterative process rather than primarily through a term-long paper. I hope this podcast episode makes you think about whether your class should have a semester-long group research project or instead focus on more targeted mini-projects.

→ ACM Code of Ethics and Professional Conduct

Great read. Give it 20 minutes. Summer can be a great time to think about these big ethical questions, ones that can't stay as abstract concepts but must become a daily practice. Here's the PDF version.

→ MAP-CS (Mission Aligned Programs for Computer Science)

A multi-institution collaboration supporting mission-aligned computer science programs at liberal arts institutions, with support from the U.S. National Science Foundation. More info here.

🦄 Quick bytes from the industry

→ Exploring with agents

I enjoyed this conversation between Adam Stacoviak and Amelia Wattenberger, designer, data visualization veteran, former GitHub Next member, and now working on Intent at Augment Code.

They discuss the idea that the last 30% of any software project may soon become the hardest part of development. Amelia argues that software engineers are experiencing an identity shift as AI agents increasingly take over implementation tasks. The conversation also explores the redesign of developer tooling in this agent-first world, the evolution from autocomplete to chat to CLI and back to UI, why Intent treats the workspace rather than the chat thread as the core primitive, the tradeoffs between one-worktree-per-agent versus one-worktree-per-task, and why prototyping has become easier while finishing products has become harder.

→ What jobs are AI jobs?

Another great conversation between Benedict Evans and Toni Cowan-Brown on Another Podcast, especially the part about the human element and why jobs like consulting aren’t really AI jobs.

Toni argues that as AI becomes increasingly powerful and eventually taken for granted, the real differentiator will shift from the technology itself to the human element behind its implementation. The real value will come from knowing how to apply AI effectively to real-world problems. Rather than replacing consulting and expertise, AI may actually make trusted human judgment even more valuable people who can turn technology into strategy, execution, and decisions tailored to specific contexts.

🌎 Computing Education Community

Last day for ITiCSE 2026 early bird registration.
ITiCSE is looking for volunteers for the Conference Committee and conference host teams.
Participants Needed for a Study on Essay Writing and ChatGPT (shared by Andrew Jelson)
UR2PhD opens summer programs for undergraduate researchers and graduate mentors, including synchronous research training courses and an asynchronous pre-research experience designed to build foundational research skills and prepare students for graduate study.
Estrella Mountain Community College is hiring a full-time Artificial Intelligence & Machine Learning faculty member for Fall 2026 (shared by Jim Nichols).
ICCE 2026 (Nov. 30–Dec. 4, Christchurch, New Zealand) is now inviting submissions across seven sub-conferences in computing and education research.
AccessComputing is hosting a book club on Digital Accessibility Ethics (May 21–June 11), focused on disability inclusion, AI, cybersecurity, design, and accessibility in tech through discussion-based sessions (shared by Brianna Blaser).
CCSC Northwestern 2026 (Oct. 9–10, Whitman College, Washington) is accepting submissions for papers, tutorials, panels, workshops, and student posters in computing and CS education. Deadline: June 30 (shared by John Stratton).

🤔 Thought(s) For You to Ponder…

David Alayón doesn’t watch traditional TV. He goes straight to specific sources and has meaningful conversations; he ends up hearing about the important stuff through other people or social media.

One issue related to gambling that I’ve discussed with many young people is self-control, especially how weak it can be among those who are constantly connected and heavily attached to their phones. Their ability to regulate impulses is still developing, which can increase the risk. Gambling is particularly appealing to this age group.

ChatGPT and Claude already “teach” languages better than they did two years ago. Duolingo has branding and gamification, but…

I noticed Jennifer A. Frey sounded a little frustrated with Tulsa in her interview on the Evangelization & Culture Podcast. I’m happy for her and for the UVA students and faculty. Great hire!

I recently discovered the work of Justin Parpan, an art director in the animation industry and a design professor at USC, who creates incredibly textured, semi-abstract digital paintings. Those color palettes are amazing! He also has a Substack where he shares the processes he uses to combine digital and analog textures to achieve the final look and feel. The artwork is outstanding on its own, so getting this level of insight into the creative process is the icing on the cake.

The Art Of Everything

How I Think About Texture

I didn’t really think about texture for a long time. I was working digitally—everything clean, controlled. Which is great, but after a while it started to feel a little flat. Like something was missing, even if I couldn’t name it…

a month ago · 75 likes · 8 comments · Justin Parpan

This study showed that AI without ever seeing the patient and using only the same clinical information a doctor receives during triage was more accurate than human physicians at determining the urgency level. Of course, AI does not replace physical or human interaction. AI cannot examine the patient or speak with family members. One potential risk is that it could become a “first opinion” that physicians accept without critically evaluating it themselves. Hospitals are already offering courses and tools focused on AI applications in medicine. Several different programs already exist, and new applications are being developed to predict things like hospital length of stay or illness severity, which is especially useful for bed management and hospital capacity planning.

I disagree with Alma Guillermoprieto (the first guest) on some aspects of her geopolitical analysis, but I really liked the metaphor she uses on the cover of “Esta improbable tierra prometida” to talk about Latin America: those people in the desert who got off a bus because it broke down, yet despite the fact that it’s probably the fifth time that bus has broken down, they still keep pushing it. Despite everything, as Alma also says in this segment, we have to keep going; we have to fix what we have and try again because it’s worth it. It’s a region that has never stopped producing beauty, creativity, art, and joy, and its people are masters of creativity. The world needs that. What happens with all those external enemies in Latin America is unfortunate, but they are still free on the inside.

Carlos Pérez Laporta is the author of Nacidos para vivir. They talk about vocation, fully enjoying life, and unity of life. They recommend The Adventure of Being Human by Piñero, Open Night by Hugo Mujica, Confessions by Saint Augustine, La Grazia by Sorrentino, and The Secret Life of Words by Coixet.

I found this interview with Fr. Mariano Fazio interesting.

Through the great books — the classics — if someone speaks to me about truth, goodness, and beauty; if they can give me the tools to distinguish good from evil, beauty from ugliness, truth from falsehood… that’s also a very natural way of transmitting the Gospel. Good literature conveys what makes the human soul come alive.
If someone jumps straight into reading The Brothers Karamazov, Crime and Punishment, or War and Peace, they may end up discouraged because those books are difficult and very long. But if you start with a book that’s more approachable, you gradually develop a taste for it.
I think we need to carry out, so to speak, an ‘apostolate’ of reading. Those of us who read should encourage young people by saying, ‘Why don’t you try this book, or that one?’ and you’ll see how it opens up new horizons for you.

Books he mentions:

MUBI Encuentros is back with a new season. They bring together two people from the film world to talk. I really liked that the first episode focused on producers, because in the end they’re the ones who decide which films actually get made. Something Marisa Fernández Armenteros said has been stuck in my head. She mentioned that she misses having metrics that could help guide her work more effectively, knowing who’s watching those films, their ages, what kinds of movies they like, and so on.

Can’t wait to listen to the new album from the Spanish band Arde Bogotá. Aimar Bretos interviewed the band in his interview segment on Hora 25.

A beautiful meditation in spanish by an Opus Dei priest reflecting on the few words spoken by Mary in the Gospel, trying to draw out possible paths and choices for our lives as Christians almost as a kind of roadmap for Christian life.

Listening to this discussion between Mila Coco and Juan Rodríguez Talavera about the work behind creating good content really resonated with me. Creating great content means constantly exposing your mind to interesting ideas. You have to consume other high-quality content that performs well, do a bit of reverse engineering to understand why it works, think about how to adapt those ideas to your own context, and figure out how to package them for each platform. It’s a continuous learning process.

Mila also made me think about the email content I consume and create. Getting attention and staying relevant in someone’s inbox has a lot to do with the relationship you build with your audience, how much they enjoy reading you, how you’ve positioned yourself, and whether you can communicate ideas in an entertaining and engaging way.

I really like the reflections in this app. It lets you create a prayer journal inspired by the daily readings.

I don’t know and it gives me food for thought:

Study conducted in January and February 2026. Coordinated by: José María Díaz-Dorronsoro. Promoted by: The Pontifical University of the Holy Cross. Sample: 9,000 young people (ages 18–29). 9 countries (Argentina, Brazil, the Philippines, Italy, Kenya, Mexico, the United Kingdom, Spain, and the United States). A redefinition of the concepts of work, well-being, and personal fulfillment among Generation Z and Millennials.

It’s true that some companies are going to need fewer employees, and that white-collar jobs (engineers, designers, strategists), and other office-based roles are among the ones being hit hardest. You can already see it happening at tech consulting firms like Capgemini or Inetum. But this situation also made me think about roles that didn’t even exist a few years ago and are now being created, like ML Researcher or AI Engineer. Maybe newly graduated professionals, building on strong core fundamentals, will increasingly need to focus on getting the most out of these agents and models: understanding the business context, figuring out how to turn models into useful tools for company data, how to fine-tune and deploy them effectively, and even how to supervise and evaluate their output. Because a lot of the time, models produce results without people fully understanding where those outputs actually come from. And companies operating in high-risk environments, for example, need that process to be tightly controlled.

📌 Research Corner

Got my spring grades back. Still surviving the PhD program.

I’ve been working hard on the ITiCSE WG. Here are a few interesting papers I read this week related to our SLR:

🪁 Leisure Line

All the best with your software engineering interviews, Nebil. I’m sure great opportunities are coming your way!

Last week, I visited Austin for the first time. A UT professor showed me around the UT Austin campus, and I wrapped up the trip with a shake at Amy’s and a visit to the Texas Capitol. Sharing a few photos from the day. The terrace belongs to my friend’s house, where we also had lunch from Pizza di Roma (highly recommend it!).

One of my weird obsessions, I guess, is being the MC at home whenever there’s a birthday celebration. It’s easy to tell which sport fits the latest birthday person. The result of the decorations in one picture. Impossible to summarize the rest of the show. It was hilarious.

First time flying with KLM, and I’ve been really impressed so far: on-time flights, great service, solid catering, and very smooth baggage coordination. The connection through AMS was super easy too. I’ll definitely fly this route again in the future.

📖📺🍿 Currently Reading, Watching, Listening

Just imagine all the auditions they must have gone through before making it to Ragtime. I love that Tiny Desk helps spotlight and give more visibility to people working in Broadway. Most of them are practically unknown to the general public, even though they have incredible talent.

What an unpleasant experience watching The Last Stop in Yuma County. I didn’t like how the story developed, but it did make me reflect on the consequences of sin.

Powerful documentary on the life of Miguel Gil on 3Cat. Some of the images he captured while covering conflicts helped change the course of history. He was killed in Sierra Leone in 2000. In the documentary, his former colleague David Guttenfelder recalls a difficult moment they experienced together in Zaire in 1997. Miguel was holding a camera in one hand while trying to help people with the other.

Madrid is already counting down the days until Pope Leo XIV’s visit in June. I was especially struck by the courageous and authentic testimony of the city’s mayor, José Luis Martínez-Almeida, in this interview with Father Ignacio Amorós and Pablo Velasco. I listened to it closely, and below I’ve gathered the passages that resonated with me most, doing my best to translate his words into English.

Reading the Gospel every day is what gives you the first step from an inherited faith to a lived faith.

From his parents he learned: consistent hard work, effort, commitment, values, passion, a sense of humor, and not taking yourself too seriously.

What AI must not do is undermine our natural intelligence: understanding, knowledge, memory — the very things that allow us to think for ourselves. I think it’s fantastic technology, but it also comes with risks.

I’ve never hidden the fact that I’m a practicing Catholic. I’m sure many of the people who vote for me are not, I’m convinced of that; but they would rather I be recognizable and transparent about who I am and what my values are than soften or disguise what I believe. I think each of us has to act in accordance with who we are, our beliefs, our intellectual background, and our spiritual baggage.

Respect is the essence of living together peacefully; it means being able to coexist despite our differences. The deterioration of coexistence begins when people start labeling and canceling those who think differently, believing they no longer have the right to express an opinion.

I love the chosen motto: “Lift up your eyes.” More than ever in the times we live in, I think it’s essential that we learn to lift our gaze — not remain fixed on the ground or attached only to a worldly perspective, but raise our eyes to understand what is truly important and transcendent, recognize it clearly, and act accordingly.

There’s going to be a Corpus Christi procession with the Holy Father through the streets of Madrid!

In terms of the city itself — its image, international standing, and global projection — it’s going to be extraordinary.

Seeing the Pope for four days, the vigil, the Mass, the events he will celebrate, the Corpus Christi procession, and a city fully devoted to this visit will be something truly exceptional. I’m sure grace will be poured out over all of Madrid. He is a moral and spiritual compass that benefits all of us, and I’m convinced this visit will bear abundant fruit.

The three pieces of advice the Pope gave him:

Be courageous. Don’t hide. Don’t lower your head. You are Catholic and you are in public life.
Never give up your principles.
The ultimate limit to any decision is respect for human life and the dignity of the person.

In Madrid there are realities that are not so visible or admirable that, were it not for the Church, Cáritas, and the Church’s charitable work, would hardly be sustained by anyone in this city. Many people want to reduce the Church to an NGO. But that charitable work is born from love for Christ. It is precisely the spiritual dimension of the Church that makes such immense service possible. And those who stand beside the most vulnerable — those whom nobody wants to see or be with — do so precisely out of love for Christ.

—What makes you happy today?
—My wife and my son.

Pablo Velasco: the best thing you can give a child is to love your wife, for him to see how you love his mother.

If you ask me what takes away my joy, I’d say that sometimes it’s not being able to make the right decision. And also making decisions knowing they may not turn out the way you hoped.

But you have to face it. I think about that too — it happens to all of us; we’re human.

What we cannot do is refuse to correct ourselves out of fear. If you make a bad decision, if you mess up, you correct it and acknowledge it. And people understand that; there’s no problem with it. You don’t have to pretend to be infallible.

💬 Quotable

In the twilight of life, we will be judged on love alone.

― St. John of the Cross

🌐 Cool things from around the internet

A collection of links to stuff I think are worth sharing.

🔗 Miniatur Wunderland — hours of craftsmanship, thousands of details.

🔗 im Water — wow, such a cool idea and design.

🔗 Sean Penn in The Secret Life of Walter Mitty — the snow leopard scene is still one of those moments that makes the world stop for a second. That reflection on living in the present, in the moment…

Issue #44 of Computing Education Things was written while listening to:

🔗 Quick Links

🎧 Listen to Computing Education Things Podcast

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

Buy Me A Coffee

#43 — Why The Demand for Systems Thinking Isn’t Going Away

Daniel Prol — Fri, 08 May 2026 10:20:54 GMT

Reflections

Photo by Alina Grubnyak on Unsplash

Programming Throwdown Is Back with an Episode on Agentic Coding

There’s a term that’s been gaining a lot of traction lately: Agentic Coding (systems that don’t just respond, but take action, executing complete processes with little to no human intervention). Major companies are already deploying these agents in real-world operations, making the topic impossible to ignore.

That’s exactly why the latest episode of Programming Throwdown is worth paying attention to. Patrick Wheeler and Jason Gauci return for a deep dive into agentic coding, where it came from, how it works under the hood, and what it really means to use these systems seriously in modern computing.

A Brief History of AI-Assisted Coding

IntelliSense has been around for decades (deterministic autocomplete based on the functions and types in your codebase). Then came GitHub Copilot, which uses LLMs to suggest entire blocks of code, not just symbol names. Cursor pushed things further by introducing multi-location tab completion: you accept a suggestion, and the tool flags follow-up work elsewhere in the file (imports to add, signatures to update…) so you can tab through a chain of related edits.

Then Claude Code arrived and changed everything. Instead of extending what you had partially written, you could describe what you wanted in plain English, and the agent would go off and do it. Jason has never written a single line of machine code in his life; he’s always run everything through a compiler. Passing instructions through an agent instead of writing code directly isn’t categorically different, he says.

As I read a few months ago from Kevin Naughton Jr.:

If you look at the history of computing, this isn’t a radical departure; it’s actually a natural progression. We started with binary and punch cards, moved to assembly, then to low-level languages like C, and eventually to high-level languages like Python and Java. Each step was about adding a layer of abstraction to make us more efficient. Communicating with computers in English is just the next logical layer.

Knowledge is King, as Kevin likes to remind us:

You must maintain an intimate understanding of how your systems work. If you don’t understand the underlying logic, you are only as good as the LLM you are leveraging. AI can handle the labor, but you must provide the oversight. Adopting the tools is mandatory, but maintaining your expertise is what makes you a 10x engineer rather than just a 10x typist.

The Software Engineer Weekly

How I Use AI to Code

With AI agents, a lot of the actual engineering work isn’t always visible to the rest of the team. Linear acts as the coordination layer for that high-velocity output, tracking work done by tools like Cursor or Claude at the issue level so you never lose visibility. It’s the first project management tool I’ve experienced that’s genuinely a joy to use…

3 months ago · 24 likes · 3 comments · Kevin Naughton Jr.

What “Agentic” Actually Means

When the model receives a question, it doesn’t just emit text, it can invoke tools: run a linter, read a file, apply a patch, or execute a Bash command. The tool’s output is added to the context, and the model uses that to decide what to do next: call another tool or finally answer.

What makes it agentic is that this loop can recurse. Jason walks through a concrete example: ask it to remove all lint errors from a codebase, and it runs the linter, sees ten thousand errors, and, rather than trying to fix everything in one enormous diff, creates subtasks. Each subtask focuses on a narrow class of error and runs as an isolated sub-question. The original model becomes an orchestrator. Each subtask runs with a smaller, more focused context, which matters because, as both hosts emphasize, model quality degrades as context grows. A model working on one chapter of a textbook produces a better summary than one trying to hold the whole book in memory at once.

The loop also makes eventual correctness possible. If the model hallucinates a method call that doesn’t exist, it can try to compile, catch the error in the tool output, and correct itself without any human intervention. Patrick notes that this is still imperfect around the edges, but the architecture is what enables recovery at all.

The Context Window Problem

Both hosts spend significant time on this because it’s the most consequential practical constraint. Every tool call result, every model response, every file chunk. All of it accumulates in the context. When the model approaches its limit, it compacts by summarizing older parts of the conversation to free up space. The problem is that summarization loses detail, and in code, details matter. Jason describes situations where, after compaction, the model repeated a question it had already asked. The thread of reasoning had been compressed away.

The mitigation is task decomposition: break large requests into smaller, focused subtasks. Each subtask runs in a fresh, narrow context. The model is smarter, faster, and less likely to go sideways. This isn’t just a technical workaround; it’s a discipline that produces better results structurally, not just computationally.

AGENTS.md and Code Hygiene

Most agentic tools look for a file called AGENTS.md in your project directory, recursing up the file tree to collect all of them. Whatever you put there gets prepended to every session. This is where you encode standing instructions: always run the unit tests after significant changes, use the virtualenv Python rather than the system one, keep files under a certain size, and break them up semantically if they grow too large.

Jason learned this last rule the hard way. He told the agent to split a large file when it got too big and ended up with cube_part1.py and cube_part2.py. He had to go back and add explicit instructions: splits must be semantically meaningful, not just mechanical. The AGENTS.md file is where you enforce good habits so you don’t have to repeat them, and where you gradually accumulate the lessons the agent would otherwise keep needing to relearn.

What Still Goes Wrong

Beyond context limits and compaction, there are still hallucinated API calls. The model assumes a method exists because the class pattern suggests it should. There’s also a tendency toward spaghetti: agents will happily duplicate logic rather than refactor it into a shared function, because deduplication requires a kind of global awareness that’s expensive in tokens. You can push back on this in AGENTS.md, but it requires active enforcement.

There’s also the question of when to push forward versus when to start over. Sometimes something in the context window pollutes the model’s reasoning, and it gets stuck in a loop. Reverting to an earlier Git commit or simply closing the session and restating the question with fresh framing often works better than trying to argue the agent out of a bad state. And since there’s randomness built into the model, temperature and other factors, you’re not guaranteed to get the same answer twice, even from the same starting point, which means a fresh attempt isn’t wasted effort.

Git, they argue, isn’t optional here. Use it even on solo projects, even if you never push to a remote. It gives the agent something to revert to, makes rollbacks cheap, and produces detailed commit messages that capture why something was done, context that won’t be inferable from the code alone.

On the Future of Software Engineering

Neither host thinks the job of software engineer disappears. The demand for systematic thinking, turning abstract problems into concrete solutions, has never declined historically, and there’s no obvious reason it would now. What changes is the nature of the work, as well as the profile and role of the software engineer. Rote porting tasks, mechanical boilerplate, the kind of work where the spec is clear and the output is predictable, are the most vulnerable. What remains is judgment: knowing what to build, recognizing when something is wrong, understanding why one architecture holds up under load while another doesn’t.

Jason makes a strong point: Salesforce isn’t just a database with a front end. It’s a decade-plus of learned lessons, role-based access control added after an intern wiped a client’s data on their last day, audit trails added after something went wrong that no one anticipated. That institutional knowledge isn’t in any GitHub comment. An agent can reproduce the surface, not the scar tissue. Companies that cancel their SaaS contracts to rebuild in-house are, in his view, setting themselves up for a round of security incidents before quietly returning.

Patrick’s framing: curiosity, taste, and the persnicketiness to keep going when things don’t match expectations. Those are the traits that will matter. The lines between disciplines may blur as engineers reach further outside their domain, but the difference between someone who builds something people actually want and someone who just ships tokens to disk will remain.

Teaching Faculty Hiring

My co-advisor, Michael Hilton, was on my favorite academic podcast, The CS-Ed Podcast, discussing a CRA memo he co-authored: Best Practices for Hiring Teaching Faculty in Research Computing Departments. What a great combination.

These ideas got me thinking:

Michael shares his own path from community college at Grossmont College, to transferring to San Diego State University, to nearly a decade as a software engineer, then returning for a master’s at Cal Poly San Luis Obispo and a PhD at Oregon State University before becoming a teaching professor at CMU.

Unlike tenure-track hiring, teaching-track processes and norms are often reinvented at every institution, forcing candidates to prepare entirely different materials for every search. Michael explains that, while tenure-track hiring benefits from widely understood community norms, teaching-focused positions often lack that shared structure, creating confusion and unnecessary burden for both candidates and departments.

From there, they walk through the hiring process step by step:

Writing job ads that don’t assume candidates can read between the lines about degree requirements, prior industry experience, contract length, long-term job security, or even what “professor of practice” is supposed to mean at a given institution.
Building evaluation rubrics before candidates arrive so mixed research-and-teaching committees align on what actually matters for the role, rather than evaluating candidates through incompatible expectations. They also discuss how rubrics reduce bias and help departments assess candidates based on the realities of the job itself.
Rethinking the job talk format so it reflects real teaching conditions rather than a faculty audience pretending to be students. Michael makes a compelling case that the traditional “teach faculty as if they were undergraduates” model often produces artificial dynamics: some faculty disengage because they already know the material, while others turn the session into a search for edge-case gotcha questions. Instead, he argues for more meta-level discussions about how candidates would actually structure and teach a course.
Structuring interview visits in ways that help both sides assess fit accurately, including letting candidates sit in on a live class so they can see what teaching at the institution actually looks like without having to prepare an entirely new lecture.

They also discuss the mismatch between fall and spring hiring cycles, which creates a particularly painful situation for teaching-track candidates navigating very different institutional timelines. Teaching-focused positions at liberal arts colleges and teaching-centered institutions often hire in the fall, while many R1 institutions conduct searches much later in the spring. As a result, candidates are frequently forced to make major career decisions before they have the chance to evaluate the full range of opportunities available to them. This becomes even more difficult for dual-academic-career couples, especially when one partner is on the traditional tenure-track market and the other is pursuing teaching-focused roles. In many cases, teaching-track offers arrive months before tenure-track interviews and offers even begin, making coordinated decision-making nearly impossible.

They also discuss the use of exploding offers as a pressure tactic. Short-deadline offers designed to force candidates into accepting positions before they can complete other interviews or compare opportunities. Michael argues that this dynamic disproportionately affects teaching-track candidates because of the fragmented timelines and lack of shared hiring norms across institutions.

One especially interesting point is how some R1 institutions are now experimenting with both fall and spring hiring cycles after repeatedly losing strong candidates who had already accepted offers elsewhere before the R1 hiring process had even started.

While I recommend reading the full memo, since it’s full of practical guidance, this episode adds the stories and nuance that make clear why getting this right matters for departments trying to find great candidates and for candidates trying to find the right fit.

I agree with Kristin that, regardless of where you are in the hiring cycle, this is an important episode. Michael goes into details that don’t appear in the memo and really highlights why it’s worth spending a little extra time to make hiring better each time around.

Quote from Michael that I liked:

I always tell my students who are going on the job market that you’re interviewing them as much as they’re interviewing you.

What skills do people need to successfully code with AI? CS + writing skills

Programming without writing a single line of code: this is the promise of vibe coding. A new study by Sverrir Thorgeirsson, Theo B. Weidmann, and Professor Zhendong Su from ETH Zürich examines which skills are needed to build simple apps with AI agents.

The results show that participants with stronger computer science backgrounds achieve better outcomes. Writing ability also has a clear impact, while frequent everyday use of language models does not improve performance.

The work highlights the role of human expertise in prompting, critical assessment, and system understanding. It shows how computer scientists shape effective collaboration between humans and AI.

Read the full article.

You can read the paper “Computer Science Achievement and Writing Skills Predict Vibe Coding Proficiency”, for all the details.

🔍 Resources for Learning CS

→ If LLMs can generate the code, what’s left for you?

In this keynote, Vicki Boykis argues the answer is everything that actually matters and none of it comes quickly: data sense, system design, and the hard-earned intuition to recognize when things are wrong.

→ The Laws of Software Engineering book is out

This new book by Milan Milanović distills more than 20 years of hard-earned software engineering lessons into 56 laws covering architecture, teams, quality, scalability, and decision-making. It connects ideas such as Gall’s Law, Brooks’ Law, Goodhart’s Law, the Two-Pizza Rule, and the Cobra Effect through real-world industry experience.

→ Allocating on the Stack

This article describes allocation optimizations in the Go compiler. Even if you don’t write Go, these detailed articles are a great read. Speaking of Go, Burak Karakan argues that Go is unusually well-suited for AI coding agents..

🔍 Resources for Teaching CS

→ Python IDEs for teaching

What IDEs are people using nowadays for introductory Python courses, especially now that tools like PyCharm have started shipping increasingly aggressive AI-assisted code completion?

This week I’ve been gathering some interesting projects from the current ecosystem of educational Python IDEs and browser-based programming environments:

TigerPython: minimalist browser-based Python IDE running fully client-side with Pyodide.
TigerPython Parser: open-source localized error-message system.
BlockPy: dual block/text Python environment with tracing, datasets, autograding, and browser execution
BlockMirror: block ↔ text synchronization editor.
CORGIS datasets for education.
Pedal autograding framework.
Drafter educational web library.
Strype: structured editor between blocks and text.
Hedy: gradual programming language for education.
Wordplay: multilingual educational programming environment.

Other tools:

There’s also an emerging tension between professional tooling and pedagogical tooling. The “best IDE for software engineers” is not necessarily the best IDE for someone learning loops, state, functions, or debugging for the first time.

Really interesting space right now for computing education, and thanks for your contributions: David Reed, Kevin Lin, Cory Bart, Tamara Nelson-Fromm, Leandro Silva Galvão de Carvalho, Manuel A. Pérez-Quiñones, Nicholas Weaver, Ken Arnold, and Neil Brown.

→ Distributed Systems lecture series

Martin Kleppmann (mentioned today in the industry section) published an awesome free Distributed Systems lecture series some time ago. If you’re teaching it in Fall 2026 or Spring 2027, it’s a great way to review some of the most important topics.

🦄 Quick bytes from the industry

→ Great episode with Martin Kleppmann

What a great episode of The Pragmatic Engineer Podcast with Martin Kleppmann, the author of the Designing Data-Intensive Applications book, academic and researcher:

Martin Kleppmann argues that academia and industry are not opposing worlds, but complementary environments with different incentives and time horizons. In academia, researchers have the freedom to pursue long-term ideas that may not be immediately commercially viable, such as his work on “local-first software,” which aims to give users more control over their own data instead of locking them into cloud platforms. He sees this as an opportunity to prioritize what is best for users rather than what maximizes business incentives.

It’s wonderful to get to work on interesting engineering and computer science problems while at the same time, like trying to pursue this higher level vision.

Kleppmann also reflects on teaching and the Cambridge system, which he describes as traditionally theoretical and heavily focused on fundamentals rather than rapidly changing technological trends. He notes that computer science education at Cambridge still emphasizes ideas that have remained relevant for decades, such as distributed systems principles and formal reasoning, instead of constantly chasing the latest industry fashions. His own teaching style mirrors this approach: lectures are often theoretical and based on careful reasoning about algorithms, failures, and trade-offs, though he also values hands-on engineering experiences in areas like cryptographic protocol implementation. He appreciates that this environment encourages students to think rigorously from first principles rather than simply applying fashionable tools uncritically.

At the same time, he values practical engineering and believes academia should stay connected to real-world problems. He argues that industry often moves too quickly and relies on trends or secondhand ideas without deeply reasoning through trade-offs, while academia encourages more rigorous and first-principles thinking. Ideally, people should move between both worlds throughout their careers, since industry experience gives researchers broader perspective, and academic training develops critical thinking that can improve engineering practice.

On AI, he sees AI tools as potentially very useful for feedback, brainstorming, and software development, but believes some forms of learning and thinking still require direct human effort. For him, writing is part of the thinking process itself, which is why he still writes his books manually. In education, he thinks AI creates a difficult challenge: universities want students to learn how to use these tools responsibly, but also need to preserve the intellectual struggle that is essential for real learning.

→ I loved the conversation with Félix Ruiz on the Tengo un Plan podcast

I recommend this episode of “Tengo un Plan” if you’re into startups and entrepreneurship. The title is a bit misleading. It’s not about easy money or get-rich-quick schemes, but about the story of Tuenti (among others). So nostalgic! And Félix Ruiz really had great vision.

🌎 Computing Education Community

Shuchi Grover (May 12): Raspberry Pi seminar on K–12 data & computing competencies and AI integration across the curriculum. Pre-register for this seminar on Zoom.
Marie Devlin seeks instructors for a 10–15 min survey on GenAI’s impact on capstone/team projects.
Amanda Holland-Minkley (MAP-CS) invites faculty to a 10-min survey on curriculum development and review processes.
The Citadel is hiring a tenure-track Assistant Professor or Instructor in Cyber Operations. Teaching-focused role with strong cyber labs, NSA CAE designation, and start dates across 2026–27 in Charleston, SC.
The Computer Graphics & Animation Research Projects (CG&ARP) program returns May 18 with a 6-week international research experience pairing undergrads, grad mentors, and faculty on graphics/animation projects.
EngageCSEdu released a special issue on game-based learning and serious games in computing education.
EduCHI is seeking expressions of interest to host EduCHI 2027 and 2028 following its addition as a SIGCHI-sponsored conference. EOIs due May 31. Shared by Olivier St-Cyr.
SIGCSE Virtual 2026: Round 1 deadline today (11:59 p.m. AoE): Panels, Papers, Special Sessions, and Working Groups.

🤔 Thought(s) For You to Ponder…

Last week, while talking with a Brazilian friend, we got into a discussion about Brazil’s Bolsa Família program. Last Friday, reflecting on the dignity of work in light of the Feast of Saint Joseph the Worker, I’ve been thinking more about government assistance. The dignity of work reminds a person that they are a child of God, with abilities, talents, and a purpose. If a government tries to support everyone primarily through subsidies, it can end up demeaning and humiliating them. There’s a difference between offering help and creating dependency. At its core, that kind of dependency can resemble a form of slavery. When a government does this, it is not truly promoting the social good; instead, it risks building a system of dependency aimed at keeping itself in power.

Anthropic is building an A+ team:

I’ve been wondering what day-to-day AI adoption looks like in companies. In my experience, getting an entire organization up to speed on any new tool or software is a huge challenge, but with AI, I’m not so sure. One question I’ve started asking friends at different companies is whether their organizations are actually integrating AI into their operations and if so, whether the benefits show up more in efficiency and productivity or in direct financial results, and whether it’s mostly being used for pilot projects or for things that have actually made it into production.

I appreciate Hadden Turner’s reminder about why we need more, not less, friction in our lives:

Over the Field

Frictionfulness

No one has ever become obese by eating too many oranges. Their peel makes sure of that. It requires frictionful effort to peel oranges, or to avoid the stones of cherries, or to scoop the goodness out of kiwis — and we moderns despise few things more than expending effort. Much easier and much more satisfying to gra…

7 months ago · 131 likes · 27 comments · Hadden Turner

We easily forget that friction isn’t always bad; in fact sometimes it is essential.
Reinstating good and necessary friction into our lives, making them frictionful in ways that direct us towards the good life rather than making them a frictionless path towards greater industrialisation, is one of the greatest needs of today.
What we need is not what we want; and what we want is not what we need.

I’d say it’s not that Claude Security can find vulnerabilities that would be impossible for a person to detect, but rather that it can uncover, at scale, issues a person could have found. Just not likely in time.

Reading Tim’s letter. Really strong, both in substance and in how it’s written.

I have no idea whether Claude Design is eating Figma’s lunch, but this article by Martin Alderson got me thinking: it’s not necessarily the company with the best standalone product, the most loyal community, or the most polished interface that wins anymore. It’s the one that controls the model, the compute, and the distribution. So could Figma end up getting absorbed into the workflow of a conversational interface like Claude?

Pepe Cruz Novillo has passed away. A legend of Spanish graphic design. RIP.

Gerard García from Deale talks with Joan Tubau about judgment and criterion in financial analysis. He points out that AI will automate a large portion of the work, but stresses that it’s still essential to know what you’re trying to achieve, to pick up on subtle nuances, and to have enough knowledge to judge what’s right and truly understand what the AI is producing. He adds that AI will make analysts more productive, allowing them to focus on higher-value tasks. At the same time, it also requires people to learn much faster. On the flip side, there’s a risk of losing some of the expertise that traditionally came from starting with a blank page or challenging assumptions. He also highlights that AI is leveling the playing field, since even small and mid-sized companies can now access advanced tools. They also discuss the situation for college students and how the way they gain experience is changing, as some traditional paths like heavy Excel work or entry-level roles are fading, even though they used to help build a professional mindset. This leads them to place more importance on curiosity in education, learning by observing others, and committing to continuous learning. They conclude that there’s no need to panic, as new types of jobs will emerge.

Kapital

K214. Gerard García. Negocios aburridos

Hay una regla en las inversiones que dice que cuanto más aburrido el negocio, mayor su rentabilidad. La gente busca las cotizadas más sexys para invertir allí su dinero pero las empresas de toda la vida siguen ofreciendo buenas oportunidades. Son negocios ocultos, en los que nadie se fija, pero que resultan esenciales para el correcto funcionamiento de …

24 days ago · Joan Tubau

I enjoyed this take from Sara Perla on praying to St. Joseph and asking him for a husband. Definitely worth a read.

The only reason I can keep trusting that Joseph lives in heaven and intercedes for me is that I have people in my life who are a bit like him. People who do things silently but who mostly keep their thoughts to themselves. They mail packages to me when they know I’ve had a rough time. They text me funny memes or offer to pick up coffee for me. Their kids start calling me “Aunt Sara” even though no one told them to. They may not share a lot of what they are thinking or feeling at a given moment — apparently, I do enough of that for all of us — but they are present. They are there. And so is St. Joseph.
St. Joseph has been silent when I have asked for his help in finding a husband, but he has not been entirely silent — he has supported me when I needed it, in ways that I did not know to ask. Just like any true friend.

It’s always a joy to hear from the Sisters of Life. This time, they teamed up with Matt Fradd to respond to real Reddit posts about crisis pregnancies and abortion. They emphasized the importance of listening first, offering practical support alongside spiritual care, and reminding women of God’s mercy and the Church’s healing resources. They also shared stories of women who found healing, reconciliation, or unexpected support after choosing life, and pointed listeners to resources like sistersoflife.org, Rachel’s Vineyard, and John Paul II Healing Center.

The profession of being an airline pilot. Ramon Vallès explains it brilliantly:

I make it a habit to say goodbye to all my passengers at the door when the flight ends. It’s my way of keeping the human side of my job alive. The best part of what I do is getting to carry people from one place to another.

My university’s Newman Center newsletter shared this talk on discerning the vocation of marriage on Sunday. You might be interested in attending if you’ll be in Houston at the end of May.

The memory shortage isn’t just threatening the smartphone and laptop industries. It’s hitting research labs too, Heidi Ledford reports for Nature. Major memory manufacturers (Samsung, SK Hynix, Micron, and others) are shifting their production toward data centers and AI chips, where profit margins are higher. Meanwhile, big tech companies building AI models are buying up massive amounts of memory for servers and GPUs.

Tools change all the time, but if you’re curious, these are the ones the Vercel design team is using right now, as shared by Hannah Hearth.

I really liked the conversation between Marta García Aller and Carlos Alsina about the importance of listening.

I loved this line from Enrique García-Máiquez: Admiration isn’t identification, but a recognition of talent and integrity.

I think, without even meaning to, one of my favorite Christian singers, Santiago Benavides, got me thinking about just how incredibly important a team is, not only in research, but in industry as well:

For me, the most important thing isn’t just playing with musicians. It’s playing with friends (who also happen to be incredible musicians). Honestly, I feel like if these songs resonate at all, it ‘s largely because they come from that deep kind of friendship.

Jorge Galindo writes about how AI is reshaping work and, more broadly, HCI here:

Rango abierto ✉️ Jorge Galindo

Claude Code: cuando el coste de cada nuevo análisis baja a cero

Mis últimos sábados por la mañana han tenido este aspecto…

3 months ago · 45 likes · 3 comments · Jorge Galindo

Fernando Arancón, Alba Leiva, and David Gómez always surprise me. Now, with an episode about the Tuareg people. A very interesting topic.

Katie McGrady has become my go-to source in English whenever something happens in the Catholic Church. Her analysis resonates with me every time I listen to or read her.

Katie McGrady

Rubio to Rome

Marco Rubio is now the Trump Administration official that has met Pope Leo the most (though Ambassador Burch has certainly been in rooms with the Holy Father most frequently), spending another half hour with him today at the Apostolic Palace in the midst of the much discussed recent “President v. Pope” controversy…

18 days ago · 24 likes · Katie Prejean McGrady

I read this article about how Alejandro Piad, whom I deeply respect, is currently using AI in his work as an academic/ researcher. In case you’d like to explore further:

The Computist Journal

How I'm Using AI Today

3 months ago · 16 likes · 11 comments · Alejandro Piad Morffis

Very nice piece by Leonardo de Moura: When AI Writes the World’s Software, Who Verifies It?

The goal is a verified software stack: open source, freely available, mathematically guaranteed correct.

Great read. Olivier Girardot:

Sometimes we want the boring stuff to stay boring.

My favorite terminal Warp is now open-source:

📌 Research Corner

I asked my students who didn’t use AI why they chose not to on an assignment that had an AI assistant built into the sidebar, and here are the patterns I noticed:

Workflow preference / low perceived difficulty: they used external tools (like VS Code) but not the in-platform AI.
Intrinsic motivation: the goal was to solve it on their own, not just get to the answer.
Independence-first approach: they would’ve used AI if they’d gotten stuck, but they didn’t need to.
Access constraint, not a deliberate choice: their independence was externally imposed.

What if your SQL query returned a chart instead of a table? ggsql adds visualization clauses directly to SQL, turning analysis and plotting into a single step. It’s still early and it was a bit of a pain to install but the workflow shift is interesting. I’ve been testing it with my current project!

I attended a seminar last Friday given by Dr. Rajmohan Rajaraman from Northeastern at UH. I left understanding a little more about the dense world of transformers. The contributions of some faculty in the room made the session even more stimulating. It was worth coming to campus on that gray and rainy day in Houston.

Waymo’s self-driving cars still rely on human support behind the scenes when the AI runs into complex or ambiguous situations, which I imagine happens pretty often in real urban environments. I like this example because it highlights that you can’t fully remove the human in the loop from the process (think grading in education).

I’ve been reading Nicholas Gardella’s PhD dissertation, Responsible and Equitable Use of AI Code Generators in Computer Science Education. It’s also available as a free audiobook. I especially liked that it explores questions of trust, well-being, and how AI is changing the experience of learning programming. It also compares human-human with human-AI pair programming in a really interesting way. Congratulations on landing a tenure-track Assistant Professor position in Computer Science at Washington and Lee University after your graduation, Nicholas!

This weekend I want to bid for SIGCSE Virtual 2026 papers. I’ll be part of the reviewers!

🪁 Leisure Line

Catching up with Minh and talking about HPC, university, AI, and life was a perfect start to the week.

Cheered on St. Thomas varsity rugby as they took on Celina in the finals at Strake Jesuit. They ended up losing, but it was a great game all the way to the end.

Last few days in Houston before heading back to Europe for the summer next week. Just enough time to catch up with Edu. Thanks for the great conversation, my friend!

Beyond the regulations, the clipping, and the whole hybrid side of things, the Miami GP was genuinely entertaining and fun to watch. It definitely improved the show. There were battles all the way to the end, position changes, and plenty of on-track action (”more driver, less software”). As for the race itself, Max Verstappen was right on the limit, but the guy is insane. He’s still firmly in the fight. It’s a joy to see the top four teams so close together. The big surprise Antonelli, who looks incredibly solid for someone with his level of experience, and Russell, who you definitely can’t count out yet. Colapinto delivered an incredible performance; also happy to see Carlos and Albon score points. Glad for Fernando too, and for both Aston Martins finishing the race actually battling on track a bit. I can’t imagine the pressure they’re under.

📖📺🍿 Currently Reading, Watching, Listening

I’d been meaning to read Confessions by St. Augustine for a while, but kept putting it off. I finally started getting into it through this audiobook, and it’s really winning me over. It feels like it came at just the right time.

A wonderful short film created by the team behind Papirola. This group of animators is on a roll right now and ranks among the very best in Spanish animation in recent years.

This song, which ties into today’s quote, got me thinking about the difference between happiness and peace: As I heard Raimo Goyarrola say not long ago, I don’t really know what happiness is; I’d rather ask: “Do you have peace in your heart? Are you living in peace?” Are you happy? I mean, think about an action movie… it ends, and that’s it, it’s over. Sure, I was happy for a moment, but is that really happiness? I’m not so sure. Now, peace. I do know what that is: living in peace.

💬 Quotable

Happiness is only experienced in moments, and you don’t realize you had it until it’s already gone.

― Gabriel García Márquez

🌐 Cool things from around the internet

A collection of links to stuff I think are worth sharing.

🔗 Isadecuppis work — few designers have the level of taste that Isabella De Cuppis does.

🔗 White Circle — control your AI.

🔗 AGENTS.md — a simple, open format for guiding coding agents.

🔗 Pi Coding Agent — a minimal terminal coding harness.

🔗 Web Rewind — an interactive journey through 30 years of the web.

🔗 Detail — your codebase is full of bugs.

🔗 p5.js — a friendly tool for learning to code and make art.

🔗 Margins — a beautifully designed reading companion for people who love books.

🔗 Landsat — your Name In Landsat.

🔗 BookRadio — it collects audiobooks that are in the public domain.

Issue #43 of Computing Education Things was written while listening to:

🔗 Quick Links

🎧 Listen to Computing Education Things Podcast

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

Buy Me A Coffee

#42 — Teaching DSA in the Age of AI

Daniel Prol — Fri, 01 May 2026 10:20:49 GMT

Reflections

In this episode, Kevin Lin (UW) and I take a broader look at the relevance of data structures and algorithms in today’s real-world problems, the role of friction in learning, and the impact of AI at different levels. We also explore online versus in-person teaching in the context of DSA, address common misconceptions, and discuss Kevin’s approach to teaching the subject. Finally, we touch on how he is reimagining computing education. This was a blast, enjoy!

Watch it on YouTube, or listen on Apple Podcasts or Spotify. Catch all the episodes here.

Mentions during the episode:

The Computing Education Things deep dives relevant to this episode:

Where to find Kevin:

CS0 and AI Literacy for Everyone in College

Image courtesy of Harvard University (public domain). Found via Houghton Library Tumblr.

Building on an idea from Titus Winters that I heard at SIGCSE TS 2026 about introducing a CS0 course for non–CS majors, not as a pipeline into CS1, but as a class focused on AI literacy, I’ve been thinking that this could actually reinforce the relevance and responsibility of those of us who have spent years studying and working in computing.

Opening up access to automation by teaching people, regardless of their field, how to take control of their computational environment makes a lot of sense, even if they understand that AI-generated code might not be suitable for professional or mission-critical use.

Of course, that level of literacy alone isn’t enough to meet the “health and hygiene” standards we expect for quality, security, and maintainability, nor will it prevent the risk of “slop.” But it feels like the minimum we should be offering if we want to stay relevant. It’s also an opportunity to give non-computing students the grounding and perspective they need to understand what the field is all about, without it sounding intimidating. It can even serve as a launching point for deeper interest if someone chooses to pursue more advanced courses.

I like how Peter Stone from UT puts it:

Nowadays, I think that, as an educator, it’s imperative that all students—regardless of discipline, whether in the humanities, fine arts, or social sciences—have at least a basic level of literacy in artificial intelligence. This helps them understand what’s realistic and evaluate headlines more critically, distinguishing between hype and what’s grounded.

The approach for computing students is different, though. As Titus suggests, it should be more grounded in fundamentals and processes. In this context, the use of AI should be carefully scaffolded, not removed, but structured in a way that pushes students beyond simply “making it work” (functionality) toward “doing it well” (quality and design).

Three main ways developers interact with code and AI-assisted tools in 2026

I like the framework Andrej Karpathy proposes for how programmers interact with code and AI tools. He breaks it down into three main categories:

The Purists: Some developers choose to reject LLMs entirely and write all their code from scratch. Karpathy suggests this is probably no longer the right approach.

Intermediate Users (the “sweet spot”): This is the approach Karpathy says he personally uses. It involves leveraging AI for specific parts of development, like autocomplete or generating boilerplate, while the developer retains full control over architecture and critical logic. AI is used to increase productivity and cognitive bandwidth, not replace decision-making.

Agent-Oriented Developers: This group leans more heavily on AI agents to handle complex tasks. However, Karpathy points out that current models are still in an in-between stage and often struggle to integrate cleanly into large or highly customized codebases, sometimes introducing unnecessary complexity.

What about vibe coding? Karpathy places this within the agent-oriented category (or as a subcategory). It’s essentially giving the model high-level prompts like “please implement this,” hitting enter, and letting it run. While he uses it for low-stakes tasks or for learning new languages like Rust, he cautions that it’s not a good fit for projects where the code structure needs to be precise and tightly controlled.

My personal take is that there’s a pretty thin line between intermediate and agent-oriented developers, since more and more like Ethan Mollick says “what AI models are best at is coding.” So experts are spending more time in an agent-oriented dev role, because most things already work, and their job is to make sure everything actually makes sense before it goes into production.

One Useful Thing

Sign of the future: GPT-5.5

I had early access to GPT-5.5, and I think it is a big deal. It is a big deal because it indicates that we are not done with the rapid improvement in AI. It is also a big deal because it is just plain good. And it is a big deal because even with all of this, the frontier of AI ability remains jagged…

a month ago · 985 likes · 74 comments · Ethan Mollick

🔍 Resources for Learning CS

→ Data Warehouse vs Data Lake vs Data Mesh

As ByteByteGo puts it: data warehouses structure data upfront for fast and reliable analytics; data lakes store raw data for flexibility (especially for ML) but require strong data governance; and data mesh decentralizes ownership so each team manages its data as a product. Most organizations end up combining all three depending on scale and use case. For further info, check out their full breakdown.

ByteByteGo Newsletter

EP212: Data Warehouse vs Data Lake vs Data Mesh

✂️ Cut your QA cycles down to minutes with QA Wolf (Sponsored…

a month ago · 305 likes · 2 comments · ByteByteGo

→ Software fundamentals matter more than ever

I really liked this talk. It pointed out that even when you’re using spec-driven development with AI, the results can still be all over the place when there isn’t a software engineer in the loop with solid programming and architectural skills. Code gets expensive fast when it’s hard to maintain or doesn’t scale.

→ Bypassing Apple’s 20% store cut (with Stripe & RevenueCat)

Example architecture: iOS handles in-app purchases and unlocks features through RevenueCat, while Stripe supports external payments by generating keys that the backend validates to grant the same permissions within the app. This is a great way to unify access.

Denislav Gavrilov

Bypassing Apple's 20% store cut (with Stripe & RevenueCat)

I made my first iOS app. While, yes, I could write about my experience using Expo, and why I think it is a great platform, this post is not about this. It is about a little technical feat I achieved, which I am proud of, and which I think can come in handy to a lot of people — precisely, one that “allows” an iOS app to accept web payments, a concept App…

a month ago · Denislav Gavrilov

→ Undergrad CS Internships

What a cool project, In the Loop! They have an episode sharing CS undergraduates’ internship stories. This podcast is based at the University of Toronto, but much of the information seems applicable beyond their school. You can listen to it here and follow their Instagram here.

🔍 Resources for Teaching CS

→ The 2026 AI Index Report

The HAI team at Stanford has just released its 2026 AI Report. Key takeaways include the continued rapid acceleration of AI development, a narrowing gap between the U.S. and China, and the U.S.’s ongoing dominance in infrastructure, alongside a heavy reliance on TSMC (the Taiwanese company) for hardware. The environmental impact is increasing, with higher energy use, CO₂ emissions, and water consumption. The effects on the job market are already being felt, particularly in junior roles. Despite record investment, there remains a shortage of talent. Around 80% of students use AI, but without clear guidelines. Differences between the most advanced models are shrinking, and adoption is becoming widespread among both companies and users.

→ How does Shazam work?

This article written by Shri Khalpada explains how Shazam works. Beyond how impressive the result is, what’s really interesting are the computing ideas behind it: signal processing, data reduction, and efficient search across large databases. A great resource for any CS educator teaching DSA or introductory concepts in signal processing.

→ Reinforcement Learning from Human Feedback

If you’re teaching LLMs, this course by Nathan Lambert is an excellent bridge between theory and practice. It builds on ideas from the RLHF Book and organizes them into a structured format, allowing post-training to be studied step by step. The course covers everything from the intuition behind RLHF to the practical implementation of key components, such as instruction fine-tuning, reward models, rejection sampling, policy gradient methods, and real-world training practices. The course covers a lot of concepts I can’t go through right now, but if you’re interested in the latest AI models, how they’re trained, and where the technology is headed, definitely check out his newsletter here:

🦄 Quick bytes from the industry

→ Mike Stonebraker interviewed by Ryan Peterman

Michael Stonebraker is part of computing history. Fascinating takes from this Turing Award winner on academia vs. industry trade-offs, how to spot real technical talent, why MapReduce and eventual consistency fall short, the cost of too many data systems, what breaks when AI agents start writing to real data systems, why LLMs struggle with real-world SQL, and how to build a career around depth, freedom, and great mentors.

Full breakdown:

He gives a clear sense of how projects like Ingres and Postgres emerged from academia (another good example is Databricks or Spark by Matei Zaharia). These are all great examples of university-grown talent that can build companies and make a real impact.
Mike says it’s “really easy” to spot smart people: just talk to them and push with deep technical questions to see how they actually think. He suggests asking things like: What was your master’s thesis? What did you specifically do? How does that work in detail? How did you handle error conditions? How many processes did you run, and why didn’t you use threads? Questions like these expose whether someone has real depth and technical rigor. As practical signals, he considers it “incredible” when someone can do three times a reasonable amount of work in school, while shallow or evasive answers are, for him, a clear sign of weak technical ability.
Mike Stonebraker argues that both MapReduce (especially Hadoop) and eventual consistency come with serious downsides. In his view, MapReduce is dramatically inefficient compared to well-designed distributed database systems because it ignores decades of database research that make large-scale analytics far more efficient. He sees its architecture as a poor fit for many data processing tasks that modern distributed databases handle much better. On eventual consistency, his criticism is more about correctness: asynchronous replication can break data integrity, leading to situations like overselling the “last item” and ending up with inconsistent or invalid states. He argues that this model doesn’t match the needs of most business applications, which depend on strong guarantees like referential integrity and accurate inventory constraints. Allowing inconsistency also forces teams to build complex compensation logic or tolerate business errors, which is rarely acceptable in critical systems. While some large-scale systems initially embraced this trade-off favoring performance over correctness he points out that even Google later moved back toward strong consistency with Spanner. His bottom line is that unless you truly don’t care about strict data correctness, eventual consistency is a risky choice for most real-world applications.
Mike argues that Amazon’s issue wasn’t a single technical flaw but an architectural and organizational one: they were running too many different database systems at once. He’s cited numbers on the order of 15 separate systems and considers that excessive. His view is that a small handful (maybe three) should be enough if chosen and designed well. Supporting that many platforms creates unnecessary operational overhead, increases maintenance costs, and makes it harder to manage expertise and consistency across teams. He also questions whether all those systems are justified. In many cases, he believes the same functionality could be delivered more efficiently on fewer, more general-purpose systems. For example, he’s been particularly skeptical of graph databases, arguing they’re rarely the most performant option and that you could often get similar results by building a graph-like interface layer on top of a relational database instead of maintaining a completely separate engine. His bottom line is that too much heterogeneity leads to complexity without proportional benefit. Fewer systems, each used well, would reduce cost and simplify operations while still covering the necessary use cases.
Mike Stonebraker has said he stayed in academia largely because it offers a level of freedom he wouldn’t have in industry. He doesn’t want a boss or to operate under corporate rules that limit what he can say, publish, or explore. In a university setting, he can openly share ideas, attend conferences, critique technologies (including those of competitors), and pursue research directions without needing approval or worrying about company messaging. He has also been put off by the bureaucracy and internal politics of large organizations. After spending time at a company like Informix, following the acquisition of the commercial version of Postgres, he found that layers of process and decision-making slowed things down and diluted impact. That environment didn’t suit his style. At the same time, academia doesn’t mean isolation from real-world impact for him. He likes pairing it with startups, where he can apply his ideas quickly and build systems in practice. That mix independence from academia combined with the hands-on execution of startups fits how he prefers to work.
Mike’s point is that agentic AI changes category the moment it stops being mostly read-only and starts performing real write operations. Today, many agents just retrieve information, make recommendations, or generate outputs without touching persistent state. But once they begin updating systems—moving money, modifying records, triggering workflows—they inherit all the classic challenges of distributed systems. At that stage, he argues, you’re no longer just building AI features—you’re effectively building a distributed database problem. That means you need strong guarantees like atomicity (either the whole action completes or nothing does), consistency (the system stays in a valid state), and proper concurrency control so multiple agents don’t step on each other’s actions. Without these, you risk very real failures, like partial transactions or conflicting updates. The implication is that most current agentic systems are getting a “free pass” by being read-only, but that won’t hold as they become more operational. Systems like DBOS are designed with this future in mind, supporting durable, transactional workflows composed of smaller steps that can still behave atomically. His broader argument is that if agentic AI is going to handle real-world actions, it needs to be built on the same rigorous foundations as high-integrity data systems.
Mike argues that LLMs currently struggle in a very practical way when it comes to working with real databases, especially data warehouses. In his own benchmarks on production-like systems, he’s seen near 0% accuracy; adding retrieval techniques like RAG might push that to around 10%, and even when you explicitly provide key pieces like the FROM clause and joins, performance only climbs to roughly 35%. His takeaway is blunt: this isn’t reliable enough for real-world use. He attributes this to a mismatch between how LLMs are trained and what real enterprise data looks like. Models are trained on corpora like The Pile, which don’t reflect the structure, naming, or quirks of actual data warehouses. Academic benchmarks like Spider or Bird also give a misleading picture—they use short, clean queries and well-structured schemas. In contrast, real queries can be 100+ lines long, and schemas are often messy, with cryptic table and column names, redundancies, and layers like materialized views. On top of that, organizations use domain-specific terminology that simply never appears in training data. The result is that text-to-SQL systems break down under real complexity. His proposed direction is more grounded: instead of expecting LLMs to magically infer complex joins across heterogeneous systems, normalize the data into tables and let traditional database optimizers do what they’re good at. He also suggests breaking complex queries into smaller pieces and giving the model more structured hints (like explicit join paths) to improve accuracy. To push this conversation forward, his team introduced the Beaver benchmark, which is based on anonymized real-world warehouses and aims to measure performance under realistic conditions rather than idealized ones.
For learning databases, Mike strongly recommends going straight to the source: seminal papers. His top pick is the “Red Book” (Readings in Database Systems), which is a curated collection of the most influential papers in the field. It’s widely used in graduate programs and gives you a solid grasp of how core ideas in databases actually evolved. More broadly, his advice is to focus on reading well-known research papers rather than relying only on textbooks. That’s where you’ll find the real technical depth and understand the trade-offs behind different systems. In his view, if you want to truly understand databases—not just use them—studying the original literature is the most effective path.
Mike is drawing a pretty sharp line between two ways of thinking about work. On one side, there’s the “job is just a job” mindset—where you clock in, clock out, and real life happens outside those hours. He’s skeptical of that framing. In his view, if you spend a huge portion of your waking life working, treating it as something separate from “life” is a missed opportunity. The alternative he’s advocating is to find work that you genuinely enjoy, so the boundary between work and life becomes less rigid and more integrated. That’s where his emphasis on passion comes in. He’s not saying “ignore reality and chase anything,” but he is pushing back against making purely economic decisions early in your career. His argument is that if you pick something you’re deeply interested in, you’re more likely to stick with it, go deeper, and ultimately do better over the long run. The people who excel technically—and end up having outsized impact—are usually the ones who are intrinsically motivated, not just optimizing for salary. He also ties this to risk-taking and differentiation. Following the crowd into whatever is currently hot or safe might feel rational, but it limits upside. Doing something slightly off the mainstream—especially in research or systems—can be riskier, but it’s also where breakthroughs and unique careers tend to come from. That idea shows up repeatedly in his own path: academia plus startups, rather than a conventional big-company trajectory. Mentorship is another lever he highlights. For grad students, he leans toward choosing the strongest program you can get into, largely because of the people you’ll be surrounded by.
For early-career engineers, the advice translates to finding environments where you can learn from people who are genuinely excellent. A strong mentor compresses years of trial-and-error into much faster growth. The personal example about his wife adds some weight to his argument. Choosing a path for practical reasons—like job security—can work on paper but still leave a lingering sense of “what if.” He uses that to underline the emotional cost of ignoring what you actually enjoy. Practically, his lens is pretty simple but uncomfortable: would you still find this work interesting if money weren’t the main driver? If the answer is consistently no, it’s a signal worth taking seriously. And beyond that, he’s nudging toward environments that let you explore, take risks, and learn quickly—because over time, that combination tends to compound into both better outcomes and a more satisfying career.

→ Great episode by Lenny interviewing Evan Spiegel

Evan Spiegel on Lenny’s podcast:

Humanity matters more than technology because adoption depends on human acceptance. That’s what will ultimately shape how AI rolls out. There’ll be a lot of social pushback if it isn’t handled well.
Tech leaders sometimes assume people will automatically adopt new technologies; Spiegel thinks that’s a mistake. He argues for putting humanity first: designing tools that move human goals forward, not just business goals.

Evan is the “non-technical” founder of Snapchat, but as a designer and artist, he has a strong sense of taste:

Design is really about empathizing with your customer, empathizing with your audience, your viewer, whoever it is, and then building something that really resonates with them.
I think removing friction from the creative process and being able to go from idea to impact—at the scale of nearly a billion people using Snapchat—is really empowering and exciting for designers, especially those who feel these tools help level the playing field for creative people.

When Snap hires designers, they focus primarily on the portfolio and look for two things: breadth (the ability to design in very different styles, which shows they’re designers, not just artists) and the story/process behind the work (why they did it and what they learned). They also value candidates with diverse backgrounds and experience moving across different products, since that brings in fresh perspectives.

Evan also explains that they’re building agents that can automate entire workflows from taking a product idea all the way through writing the spec, identifying risks (legal, trust & safety), preparing launch materials (blogs, etc.), and even generating visuals, all in a single process. If a task is well-defined enough as a “job to be done,” you can hand it off to an agent and see a significant boost in efficiency.

🌎 Computing Education Community

EduCHI 2026 (May 20–22, Toronto, hybrid) program is now live. Register by May 1 to secure early rates.
SIGCSE Virtual 2026 - Call for Submissions (Paper Abstracts due May 1).
ACM AI Leadership Summit Doctoral Consortium (Aug 2026) invites PhD students to apply by May 9.
OCCTIVE is still seeking faculty across disciplines to join its May workshop or adopt its free computing-concept video library. Apply by May 1.
CompEd seeks a Program Chair (Junior 2027 São Paulo → Senior 2029). Submit a letter of interest and CV to Alison Clear: Aclear@eit.ac.nz
Mahsa Sadeghi (msadeghi@DAWSONCOLLEGE.QC.CA) and Eric Mayhew seek instructors for a 30-min interview study on how databases are taught.
We’re hiring Professor of Practice for our Katy, TX campus.
David H. Smith IV seeks database instructors for a 60-min study on ER/schema diagram teaching (with $40 compensation). Sign up to share your approach.
Andy Zaidman shares an Assistant Professor opening at Delft University of Technology (Netherlands, hybrid).
Oana Andrei announces SCHIER 2026 (Sept 7–11, Glasgow): a 5-day summer school on Informatics Education Research for PhDs/postdocs. Apply by June 26.
Katharine Childs seeks UK primary teachers using micro:bit, Crumble, or Pico for a study on physical computing. Apply by May 31.

🤔 Thought(s) For You to Ponder…

100% agree with Fred Martin: Students who work hard still get great jobs in computing.

In the era of LLMs and all the “slop,” Hans Laguna writes this book about authenticity (everything that comes with it) as a key asset for pop stars, alongside talent (craft), hard work, consistency, self-confidence, believing in what you do, passion, and respect for the audience. It got me thinking about what it really means to have an authentic, genuine personality.

It also seems like authenticity plays a key role in political communication: it makes us more willing to listen and pay attention. We reward politicians who’ve built a message we believe in and that, in turn, helps them set the agenda. It’s not just what’s being said; it’s whether the message feels consistent with the person, their story, and who they are. Pepe Mujica is a great example. His impact doesn’t come only from the depth of his ideas, but from a life that makes his message credible. That authenticity lets him “tame the algorithm”, stand out even in a digital environment that rewards the superficial, because authenticity can be disruptive and capture attention. Closeness creates that disruption. Julián Kanarek, political consultant and author of Omitir intro.

Wonderful testimony by Matt Daniels! Another great episode of the Journey Home!

I like that A Vivir keeps investing in this segment with Pere Estupinyà and Javier Sampedro. I was especially interested in the discussion on robotics, particularly the challenge of giving machines a sense of touch and real-world awareness, which seems key for humanoid robots and the piece on Anthropic’s new Mythos model, which highlights both the immense potential and the serious security risks of increasingly powerful AI systems.

This episode of New Fire Presenta really stood out to me because it centers on what we actually mean by success. It goes beyond money or recognition and invites us to look at the heart (our intentions, our attachments, and the fruits of what we do). What I’m taking with me is that idea of seeking “God’s applause” at the end of the day, when I lie down and close my eyes, along with this deeper benchmark: if my professional and personal path is helping me love more and serve others better, I’m on the right track. If not, it might be time to recalibrate.

Recorded by Edward Sri at Catherine’s house in Siena: persevere and stay faithful, even when your mind wanders. Don’t get discouraged. Avoid distractions, and be wary of discouraging thoughts from the devil. Maintain a sincere disposition, and approach prayer with humility by offering God your nothingness. God values the genuine intention of the heart, not appearances. It’s about the will, not just feelings.

If you don’t have plans for the summer yet, this In Altum program in Washington looks really interesting. You can take part in thoughtful discussions on culture, public life, and faith, and build great friendships along the way. I’d encourage you to check it out.

The growth on GitHub is insane

@GergelyOrosz The graphs reflect growth from Jan ’23 to March ’26: 25M → 90M merged PRs/month, 389M → 1.4B commits/month, and 6.9M → 20M net new repos/month.\n\nThe intent was to highlight the relative scale between these metrics. Appreciate the feedback and we’ll factor it into future","username":"mariorod1","name":"Mario Rodriguez","profile_image_url":"https://pbs.substack.com/profile_images/1783313134117376000/hWT1AJWp_normal.jpg","date":"2026-04-28T22:10:15.000Z","photos":[],"quoted_tweet":{},"reply_count":1,"retweet_count":1,"like_count":32,"impression_count":13132,"expanded_url":null,"video_url":null,"belowTheFold":true}" data-component-name="Twitter2ToDOM">

I hadn’t listened to this interview podcast by comedian Raúl Pérez in a while, the guests just hadn’t really caught my attention, but this week Edu Soto pulled me back in.

For 14 years, documentary filmmaker Andrés Ruiz followed Colombian families in both rural and urban areas to understand how they make ends meet. The result is this incredible documentary, produced by Séptima Films in collaboration with Universidad de los Andes. If I lived in Colombia, I’d go see it. Thanks, Andrés Acevedo, for putting it on my radar.

It was amazing to discover that there’s a bookstore in Tokyo that focuses on selling just one book at a time. The idea is to encourage mindfulness and offer a range of elements—photos, objects, and printed materials—all stemming from that single title. It’s not about choosing or even buying, but simply being.

It’s a bit rough around the edges, but what Víctor Correal says in this episode is pretty interesting. Humans change depending on the tools they use. He’s not just talking about a shift in how he works, but also in how he thinks, almost like a new cognitive system, referencing Andy Clark’s idea of the extended mind. It’s also interesting what he says about McLuhan: everything we gain is something we stop using, tools as prosthetics, something you effectively “amputate.” He also mentions that what he’s willing to take on has changed, and that adoption looks different if you’ve spent years wrestling with software projects (you develop a certain judgment/criterion) versus coming in without that background, which requires a lot more curiosity.

Interesting take by Alex Imas:

The economics of structural change tells us that when technology makes one type of production cheap, the economy doesn’t collapse. It transforms. It shifts toward the things that technology can’t make cheap. For AI, those things are exactly the ones where human involvement carries inherent, irreplaceable value.

Ghosts of Electricity

What will be scarce?

Starbucks is a huge company (market cap of $112 billion) that sells one of the most standardized products in the modern economy. Making a cup of coffee or even one of the fancy specialty drinks is very easy to mechanize and reproduce. If the entire economy is soon to be automated, with labor being replaced with increasingly more sophisticated capital, S…

a month ago · 1218 likes · 129 comments · Alex Imas

Did you see the interior design of Ferrari’s first fully electric car? An absolutely stunning blend of physical and tactile elements. Designed by Jony Ive through his creative firm LoveFrom, in collaboration with Marc Newson.

📌 Research Corner

I met Max Sölch in Koli back in 2024. A few weeks ago, he presented this paper at ICSE 2026 in Rio de Janeiro. His work on assessments is always a source of inspiration and reading this one has been especially helpful for my current project.

The committee for the first edition of the University of Houston Conference on Computer Science worked hard over the past few weeks to ensure everything ran smoothly on Wednesday. I served as Program Chair, and I hope attendees enjoyed the papers selected for the conference!

This is the poster I presented at the conference. It’s still a work in progress, but I already have some preliminary results, and the system architecture is complete. I just haven’t defined the research questions yet... By the way, I’m very honored to have received the Best Poster Award.

🪁 Leisure Line

Had a great time talking with Luis on Tuesday, with the Houston skyline in the background. I learned a lot about Brazil’s healthcare system. Hope to see you next time in Fortaleza.

F1 is back! Really looking forward to Miami and what seems like a more “old-school” kind of race, with plenty of new elements on track: less “super clipping,” less reliance on the electric side, smaller speed gaps so there’s more wheel-to-wheel action and a tighter, more compact race overall. Love it. Let’s see if these upgrades spice up the championship. Hopefully Miami also brings back more of the human and sporting side of the story.

📖📺🍿 Currently Reading, Watching, Listening

When AI floods everything, craftsmanship makes a comeback: the BBC’s intro for the Milan Cortina Winter Olympics is handmade and it’s a joy to watch:

Saw Project Hail Mary this week. Beautiful and surprisingly powerful. I had such a great time with it, and it genuinely made me laugh quite a bit. It feels like old-school sci-fi: an adorable alien, a wonderfully charismatic Ryan Gosling, and I’ve heard it stays very true to the book (haven’t read it myself). And Sandra Hüller has a fantastic karaoke moment.

A year since Pope Francis returned to the Father on that Monday of the Octave of Easter. Here’s a visual narrative of his pontificate, with English subtitles, created by Vatican News, Radio Vaticana, and L’Osservatore Romano:

I’m a big fan of Isaac French. He lives in Waco (Central Texas) with his wife and two kids. He’s building Live Oak Lake (those cabins are beautiful) and living a simple life in a faith-rooted community. Recently, he picked up his new camera and filmed a full day of his life. If you’re curious about their lifestyle (work, raising a family on a farm, faith, and the “why” behind it all) this is for you. I recommend watching it when you have time to sit through the whole thing. It’s just over 30 minutes, and it’s worth it.

On Twenty Thousand Hertz, Dallas Taylor interviews Jad Abumrad (Radiolab, Fela Kuti: Fear No Man). After listening to the episode, I watched the companion video piece. Really cool if you want to geek out over his Pro Tools session.

There are days, though, when I’m trying to make a really specific sound work, and I think, ‘This is so stupid.’ But then, if you get it right, and four years later—after you’ve forgotten about it—someone encounters that thing and has an emotional reaction, you’re like, ‘Okay, that’s why.’

💬 Quotable

You have made us for yourself, O Lord, and our heart is restless until it rests in you.

― St. Augustine’s Confessions

🌐 Cool things from around the internet

A collection of links to stuff I think are worth sharing.

🔗 ggsql — a grammar of graphics for SQL.

🔗 Train Jazz — a website that turns the real time New York subway into jazz music.

🔗 Macfolio — cool website for Mac users.

🔗 transitions.dev — great animations for web development and apps.

🔗 The Public Domain Review — a website that curates fascinating gems from the history of art, literature, and ideas. It features books, films, and audio, but what I like most is its image archive.

Issue #42 of Computing Education Things was written while listening to:

🔗 Quick Links

🎧 Listen to Computing Education Things Podcast

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

Buy Me A Coffee

#41 — Research problems now matter to everyone

Daniel Prol — Fri, 24 Apr 2026 10:21:54 GMT

Reflection

This week’s title and subtitle might be confusing without context, but I’ve been wanting to reflect on this topic for a long time. David Vazquez’s words, he works across both academia and industry, really gave me the push to finally do it.

The main idea: the problems researchers used to find fun to work on, basically science fiction, now matter to everyone.

David talks about “passion in the eyes.” That’s what he looks for when hiring a new researcher at ServiceNow.

A student says, “I’ve done this and that, and it worked, and so on. And now I want to do this, and then that. And I read this paper…”

It got me thinking that in industry, passion might be in the why, while in research it’s in the how. And David puts it really well:

If there’s a problem out there that can’t be solved, I’m going to solve it. Why? It doesn’t matter if no one cares about it. I’m going to solve it.

And that brings me to this week’s title. The “sci-fi” research problems that used to be fun for us to work on now matter to people. There’s a lot of very granular interest in the AI space right now.

David gives an example of a problem that felt incredibly interesting in their lab: how a computer can learn and how to make it learn better. At the time, he says they were focused on investigating stuff no one cared about and now suddenly everyone cares.

Of course, he also makes it clear that just because these problems matter now doesn’t mean companies or universities are throwing money at them blindly. You still have to convince people, explain why now… And this connects to what David says about research teams: they can be small, but the productivity per person can be very high, usually measured by the number of accepted papers, citations, and overall impact. Some companies are seeing the value in these teams even without immediate financial returns. In the end, universities produce talent, and that talent can go on to build startups, as David says. This allows teams not only to focus on customer impact, but also on publishable, and novel work. David says it’s a balance at ServiceNow:

Something that’s useful for the company, useful for science, and novel. Because if it’s not very novel, then what’s the point of putting a research team on it? If an engineering team could build it in a few months but just doesn’t have the time right now, then it’s not that interesting.

Although he admits it’s getting harder to find truly novel problems, and more often it’s about new applications.

This is what differentiates them from research at OpenAI, which publishes a paper from time to time with a new model but is heavily focused on the product getting it to market as quickly as possible and shipping things into the product.

Another interesting point is projects that are only possible in research and would never become reality inside a company. He mentions that he recently became an adjunct professor and is now working on a project translating indigenous languages. This brings us back to the idea of contributing to human knowledge and having a social impact on more marginalized communities. There need to be people pushing knowledge forward.

To wrap up, I also found his take on why we publish papers interesting. It might sound basic, but as a PhD student, it was helpful to revisit the fundamentals:

Rigor: for him, a paper is a way to bring rigor to his work. When you have to explain something in eight pages, in detail, so others can understand it and so external reviewers who are experts in the field can read it, give feedback, and help improve it it filters what makes sense and what doesn’t, and gives the work a certain level of credibility.
Open science: so others can build on it. When people take your work as a foundation and improve it, science moves forward much faster.
Recognition: on top of that, there’s validation from other researchers that your work is valuable.

Final reflection

All in all, a fascinating episode and very different from the usual topics covered on itnig. It’s great when they occasionally bring in people like David, even though the focus is usually on founders chasing market opportunities, solving customer problems, and capturing value. The world of research is very different. It’s a world that pursues knowledge for its own sake and that’s why David was such a great discovery and deserved this reflection.

🔍 Resources for Learning CS

→ Claude Code deep dive

Someone even went and built a dedicated website to sort this out.

🔍 Resources for Teaching CS

→ Making AI education accessible to everyone

While listening to Peter Stone (UT Austin CS) on ACM ByteCast, I came across this course they created to teach AI literacy more broadly by offering an essentials course for non-technical students and lifelong learners. It doesn’t require any programming and covers history, key paradigms, and societal impacts, and it’s freely available online.

→ The Math of Weighting Proctored and Unproctored Grades

This blog post by Austin Cory Bart got me thinking about what a final grade means through the lens of how much is (and isn’t) proctored. If we want grades to represent mastery, that split has more implications than we thought.

🦄 Quick bytes from the industry

I’m sharing this link with you without much commentary because, honestly, I haven’t had the time to watch it yet. I really value your time and the space you give me in your inbox, and I didn’t want to send something half-baked or without the depth it deserves. If it resonates with me, I’ll share some thoughts with you next week.

🌎 Computing Education Community

Tommaso Carraro seeks faculty input (10–15 min survey) on SEP competencies in computing education.
OCCTIVE invites faculty to a May workshop to integrate computing-concept videos into their courses. Apply by May 1. Shared by Valerie Barr, Mia Minnes, Andrea Tartaro, Madalene Spezialetti, Kristina Striegnitz.
CCSC:MW 2026 (Oct 9–10, Kenyon College, Ohio) invites papers, panels, and workshops. Submit by May 11.
Koli Calling 2026 (Nov 5–8, Finland) is now open for submissions.
Reminder: ICER 2026 Lightning Talks & Posters deadline is today.
Anthropic STEM Fellows Program (June–Sept, San Francisco). Applications due May 15. Best suited for late-stage PhD students, postdocs, or equivalent researchers with strong domain expertise. Full-time, in-person for ~3 months, with early-stage PhDs less competitive and undergrads/early Master’s not the target audience.
Harvard & UW study seeks K–12 CS educators in the U.S. for a 30-min interview on HCI in CS classrooms. $25 gift card for participation (shared by Ritesh Kanchi).
A4NE’26 (SIGCOMM Education Workshop, Aug 17, hybrid) invites papers on networking education for the AI era. Submit by June 1 (shared by Ranysha Ware).
SIGCSE Henry Walker Travel Grants now open for RESPECT, ITiCSE, and ICER 2026. Apply by May 8 (June 1 for ICER) to support conference attendance (shared by Ethel Tshukudu).

🤔 Thought(s) For You to Ponder…

There are several people who have really had me thinking a lot this week:

I have a deep devotion to St. John Henry Newman, so this surprising essay about him and... Easter chocolate? by Dixie Dillon Lane’s husband was a fantastic read: “…this witness of mutual charity was one more step in the journey of the remarkable Cadbury family, whose frequent contemplation was the query, “And who is my neighbor?”

The Hollow

John Henry Newman, Local Hero

In celebration of the delights of leftover Easter chocolates, historian Christopher Lane (yep, my better half) and Hearth & Field have given me permission to post an excerpt from this week’s excellent essay on John Henry Newman and the virtues of neighborliness and local living…

2 months ago · 20 likes · 7 comments · Dixie Dillon Lane

Great story on CNN. Really inspiring. Gabriela Salas is an Indigenous Mexican who used her programming skills to bring Nahuatl to Google Translate.

I agree that “criterion” is a better word than “taste,” although my thinking has been evolving. In issue #20 of this newsletter, I talk about taste because it’s a concept widely used across the industry. I really enjoyed this piece. I especially appreciated how Aadarsh unpacked “taste.” Where the industry tends to say “taste,” I’ve been trying to sharpen that into “criterion” as a way to make the concept more actionable. To me, the key distinction is that taste describes the outcome, whereas criterion points to the inner process that makes that outcome possible. It’s the ability to distinguish what is actually good in a concrete situation not just what works or what looks convincing, but what is right given the context, the constraints, the people involved... It’s hard to define, but this kind of “sound judgment” feels like expert intuition, a sense of “this works,” a trained eye for quality. In terms of imitation, I agree. You can’t reach a point where you stop copying without beginning to disagree, and that’s where your own criterion starts to form. It’s about designing for the conditions of judgment, not just the production of artifacts. The open question for me is how explicit we can (or should) make that process for students.

The IdealLab

AICSEPAR 2026 Trip Report: Where Is the Pedagogy, People?

TL;AI - GenAI in CS education research seems to be chasing AI tool capabilities instead of asking the more durable pedagogical question: what conditions actually produce learning…

2 months ago · 4 likes · 6 comments · Aadarsh

Pope’s journey in Africa: Sometimes the environment doesn’t help. Let’s do our part and do it well. It’s about remembering that line from Saint John of the Cross: where there is no love, put love, and you will find love.

Great talk by Andy Matuschak! Coding agents might help us finally break out of two cages: the app model, which traps computing in one-size-fits-all silos; and programming as a specialization, which has crowded out cultures of imagination and domain insight.

Peter Steinberger on building OpenClaw at TED 2026.

Claude Code is addictive. It’s well explained here. I also think that simply working in the terminal makes developers perceive the process as more technical.

Since we’re talking about Anthropic, Erik Schluntz, a member of their technical staff, gave a 30-minute talk on how Anthropic uses AI to write code.

It’s official.

It pains us to “cut things off completely,” but many times what’s truly healthy is exactly that—the cut. — Raúl Hernández

Choosing is growing. It’s deciding who we want to be. Cutting ties with what enslaves us and hurts us so we can fly and be free. Letting go of false securities and daring to step into relationships that help us grow. Loving means letting go.

Miguel Ángel Martínez-González makes a strong, evidence-based case in the public sphere on issues like pornography, screen overuse, and the consequences of unchecked sexuality. If you understand Spanish, I recommend listening to him on Aladetres:

Marc Brooker from AWS is the kind of person who really dives into the details: postmortems, caching, that sort of thing but he also zooms out and talks about bigger topics like finding problems that actually matter, the impact of AI (on jobs, juniors vs. seniors), the benefits of writing, and even recommendations for technical books (they’re all here).

This episode with Stephany Oliveros touches on a bit of everything. It works well as a high-level overview, with broad strokes on human–AI collaboration, relevant research studies, AI’s strengths and weaknesses at work, new skills and more.

It makes sense to feel worn out by the end of the day. These are the kinds of things that fill us up when we give ourselves to others. It’s simply part of the demands of any calling. Jesus cared about His apostles getting rest. The work we do can feel nonstop—we’re always busy—but just as Jesus led His disciples to step away and rest, we need that too. It’s necessary and important. Rest isn’t about doing nothing; it’s about a change of pace, resting with others and nurturing your inner life: What gives your soul life? Where do you find peace?

I think having routines really helps. There’s a Spanish Substack that shares people’s routines.

More on the “return to the human.” Jordi González on Uri Sabat’s La Fórmula del Éxito, makes an interesting point: traditional TV isn’t going away because it can keep people company: the pure presence (mistakes, authenticity, that sense of being there with someone).

He argues that live formats (news, entertainment, etc.) aren’t going anywhere and are actually less replaceable than scripted content, which people now mostly watch on streaming platforms. He also makes a strong point about AI: it may have read everything, but it hasn’t lived anything—so it can’t truly feel empathy. That’s why human communication will still matter, even as AI keeps advancing. He also mentions that it’s getting harder and harder to tell what’s real and what’s AI-generated.

Excited to watch this. I’m already on the list!

GPT-5.5 is here. Ethan broke down all its new features.

One Useful Thing

Sign of the future: GPT-5.5

a month ago · 582 likes · 34 comments · Ethan Mollick

GPT-5.5 is clearly not the end of this process, but it is a noteworthy step along the way. I have been writing this newsletter for over three years now, and the pattern has not changed: every few months a new model arrives. I run my tests and something that was impossible becomes easy, while the size of the leaps grows each new release cycle. The jagged frontier is still there. It is just much further out than it used to be.

📌 Research Corner

Yesterday, I was invited to give a session on AI and spark a discussion about where it’s heading and how it affects us at different levels. This time, I focused on the workplace and the human side as well. I really enjoyed interacting with the small groups!

This week, I released the final assignment of the spring semester for my Software Design class, and I’m now reviewing submissions. I gave students three design pattern problems we covered in class and allowed complete freedom in using AI. No restrictions. I wanted to observe their chain of thought to better understand their thinking, judgment, and decision-making. The pilot phase of my project at UH has now wrapped up. After focusing primarily on the development side, I’m shifting more toward the research side. I’m also excited to try my new tool at Aalto this summer.

I stopped by the Undergraduate Research Showcase for a bit last Wednesday. One of the things I tell my Computer Science students most often is that presenting posters is valuable. Along with hackathons and coding clubs, it’s one of the experiences that makes the biggest difference.

I liked this poster. The student created a game in Unity to educate construction workers on cybersecurity issues and help prevent more serious problems.

I really liked the graphs on this poster too. Jacob did a great job explaining his research on distributed algorithms and the APSP problem in distributed networks.

🪁 Leisure Line

On Friday evening, I had dinner with my friend Cayo. Always a treat! I’ve heard that the weekend brunch at this place near my home in Houston called Canyon Creek is delicious.

People who are alone sometimes just need a hug. They need someone to be with them, to show them love, to meet them where they are in their reality, like a mother who kneels down to her child, the way God does with us.

When we go to a nursing home, we can also be that embrace of God for many people. That kind of presence can change hearts. Hopefully, this Easter, we can open our own hearts as well allowing God to meet us, letting ourselves be embraced by Him, so that we can then go out, meet others where they are, and be that embrace of God for them.

📖📺🍿 Currently Reading, Watching, Listening

The pilot of Gnome Show is now on YouTube. It’s created by Cas van de Pol, and the animation style is fantastic. Not to mention the character design. Special shoutout to the backgrounds, which are absolutely gorgeous.

The Missing Post Office is a place in Japan where letters are sent that will never reach their recipients (sometimes because they’re no longer around). This short documentary is a beautiful work of art, in case you’d like to see more:

💬 Quotable

Love isn’t just one happy moment, right? It’s a million. And it’s bickering in the car, and supporting someone when they need it, and it’s growing together, and looking after each other.

― Eternity (2025)

🌐 Cool things from around the internet

A collection of links to stuff I think are worth sharing.

🔗 Fractal Paris — it brought back some beautiful memories. Lovely!

🔗 Claude Design — it generates interactive prototypes, presentations, and other visual work just by chatting with Claude.

🔗 So cute… This Instagram reel by Jonathan Romero really hit me.

🔗 The Sound of Love — it plays a romantic song every time you click on the page and shows a comment someone left on that song’s YouTube video.

Issue #41 of Computing Education Things was written while listening to:

🔗 Quick Links

🎧 Listen to Computing Education Things Podcast

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

Buy Me A Coffee

#40 — How Do CS Departments Support Their Teaching?

Daniel Prol — Fri, 17 Apr 2026 14:04:40 GMT

Reflections

In this episode, I spoke with Borja Sotomayor, a Senior Instructional Professor in the Department of Computer Science at the University of Chicago and former Director of the MPCS, about how computer science departments support their teaching activities. We explored topics such as balancing teaching and research priorities, managing large classes, onboarding new faculty, the role of mentorship, evaluating teaching effectiveness, and the need for more structured support for CS educators.

I really enjoyed my conversation with Borja. Watch it on YouTube, or listen on Apple Podcasts or Spotify. Catch all the episodes here.

Interesting quotes from the episode:

I think it’s important to be aware of what students will need to do in industry, but I don’t think industry should dictate what we teach in college.
Co-teaching is a really effective way to learn the ropes of how things are done at an institution before teaching entirely on your own.
Mentorship is great in principle, but it also has to be very deliberate. It can’t just be a department saying, “We’ve assigned you a mentor,” and then hoping that everything will fall into place just because those two people meet periodically.
If someone is going to be a mentor, they should be trained on what it means to be one, how to be effective in that role, and how to provide that kind of support.
Teaching a class involves much more than just showing up for lecture, speaking for a set amount of time, and being done. That might be true for small or even very small classes. But when you’re running a large class, you’re effectively managing a staff: a team of TAs and graders.
When you have a very large class and a large team of TAs, you have to start thinking about structure. For example, you might have a few head TAs who manage the graders, so you don’t have to oversee every individual member of the course staff directly. Some universities also have roles sometimes within instructional support staff such as course assistants or course coordinators, who help manage these responsibilities. This is something I was never explicitly taught, and many people aren’t. I think part of addressing that gap is ensuring that, at the PhD level, there is more training in pedagogy.
At many institutions, especially for tenure-line faculty, the only thing that’s considered is student evaluations. There’s a substantial body of literature showing that these are quite problematic. They can be useful in certain contexts, but not as a comprehensive measure of teaching effectiveness.

Mentions during the episode:

The Computing Education Things deep dives relevant to this episode:

Where to find Borja:

How to develop criterion

Today’s reflection might sound a bit philosophical or abstract, but bear with me, it does connect to the topics I’ve been exploring in this newsletter.

“Criterion” comes from the Greek krínein (κρίνω): to separate, to distinguish, to keep the wheat and discard the chaff. Criterion is not about “going by the book” as a form of blind obedience. It is the inner ability to clearly see what is good in a specific situation without being clouded by frustration, resentment, or the need to be right.

To have criterion, we need a well-formed conscience that is clean enough to look through, like a dust-free window. A conscience that does not lie to itself, is not dragged around by emotion, and can weigh pros and cons without the scale already being biased from within. That is what makes someone a person of criterion: someone who knows what to do with what they know.

AI has democratized execution. Execution is no longer the barrier. What AI has not democratized, however, is discernment: the ability to tell whether what it produces is genuinely good or merely looks good. The real bottleneck lies in distinguishing what works from what is right, and in asking ourselves, before delivering anything, whether it reflects what we truly think or simply what was easiest to generate.

In computing education, a student without criterion can rely on AI to produce work without ever truly understanding. It may look convincing for a time until the moment comes to justify it: Is this good? Why? What do you think? Without criterion, there’s nothing to say. Just output.

The problem is that criterion cannot really be taught directly. It has to be cultivated, or more accurately, we have to allow ourselves to be cultivated by it. Now think about this with Holy Week still fresh in our minds: when Jesus chose His apostles, He knew that enthusiasm and emotion weren’t enough. He didn’t rely on words alone. He formed them through concrete situations, each one different, so that those lessons would gradually take root in their hearts. The washing of the feet at the Last Supper wasn’t just symbolic or theatrical. It was formation by example, lived out in a tangible way.

There’s no single method, but these four ideas can help us develop stronger criterion:

The first is friction. Criterion does not enter through our eyes; it is shaped through contact, like one stone polishing another. That is why it matters to surround ourselves with people who are better than us, people who challenge us, who do not simply confirm what we already think. If it feels uncomfortable, we are probably in the right place.

The second is struggle. The kind that stings, that leaves a mark. Criterion needs scars, not theories. A professor who does not allow real failure is not forming criterion. Forming criterion requires putting all our heart, will, and inventiveness into the process by a disposition of service, to offer those students around us a dialogue of real light. Criterion helps them to be freer, and this is especially important for young people.

The third is reading and speaking with people. Not reading to know more, but reading in a way that unsettles what we already believe. If every paper, book, or article reinforces our current views, we are reading wrong. Criterion needs cracks to enter through. Those cracks open when we expose ourselves to what does not fit our framework, to people who think differently, to conversations whose outcome we cannot predict.

The fourth is copying. Imitate those we admire. Not to remain imitators, but because there comes a point when we can no longer copy without betraying something inside us. That moment matters. That is where something truly ours begins.

Distrust quick answers, especially our own. Criterion is rarely fast. Pope Pius X, when asked a question, would say: “Let me think about it first.” He would think, pray, and then respond, not always in that order, but never skipping any of the steps. That too is criterion: recognizing that some questions do not deserve an immediate answer, and that the urgency to respond quickly is often more about ego than clarity.

But criterion without action is just contemplation. The question before any task is simple: is this going to be an act of genuine work, of real service? If yes, do it. And if we feel tired or anxious, that is not a reason to stop. It is an invitation to reframe. From “I don’t feel like this” to “I am going to do this well, for someone.” That shift, small as it sounds, is criterion in motion.

It is worth cultivating criterion. Because in a moment where production is infinite, the one thing we cannot generate with a prompt is the ability to judge what has been produced. And criterion will never be entirely ours. It will be, at best, a trace left by all the interactions, errors, and encounters that have shaped us along the way.

Cultivate it. Or better yet, let ourselves be cultivated by it.

Using Two-Stage Exams to Promote Active Learning in Large Classes

I found this episode with Notre Dame’s Rachel Branco really interesting and relevant to our ITiCSE 2026 working group on teamwork and collaboration. It focuses on two-stage exams for large classes. While this approach may not be for everyone, I especially appreciated the level of detail about the context and how they implement it. It helped me better understand whether such an idea would work for us here at UofH.

The main idea is that two-stage exams increase student collaboration by having students first complete the exam individually and then immediately retake it in randomly assigned groups of three. In these groups, they explain and defend their answers to one another.

A particularly relevant point for me is how students reported feeling more connected, making friends, and developing norms of mutual support. They also felt more accountable, as social pressure discouraged them from being the weakest link, motivating deeper preparation and clearer articulation of their reasoning. In a way, students become like mini-tutors for one another.

In terms of group size, they recommend groups of three (a “sweet spot”), as this ensures everyone has a voice while keeping the group resilient if someone is absent. Groups that are too small (two) risk failure if one person is absent or disengaged, while larger groups (five or more) tend to reduce participation.

What got me thinking about the two stage exam is that it seems to address a problem very similar to the one I am trying to solve in my current research, just from a different angle. At its core, it is also a feedback timing issue. Students do not know where they went wrong until days later, when the cognitive moment has already passed. This professor addresses that through a social mechanism, with the group providing immediate feedback. We are building a system with real time tracing, so feedback does not have to be delayed.

The students who benefit most from the two stage format are lower performing students and those from underrepresented groups. That suggests the value is not just in the feedback itself, but in having someone—a peer or a TA, for example—step in before the student disengages or ends up in a group that is not providing meaningful feedback.

🔍 Resources for Learning CS

→ What is MCP?

All the docs about this open-source standard that allows AI models to connect to tools and data (such as APIs, databases, and files) via a client–server architecture, enabling structured access through prompts, resources, and callable tools without requiring custom integrations.

→ What makes Claude Mythos so important?

It can write moderately sophisticated multi-step exploits in a matter of hours, it’s a more “obedient” model to human instructions, and its alignment closely matches what’s being asked of it. Large companies and governments need security experts before Mythos is released… This is a great overview:

Understanding AI

Why Anthropic believes its latest model is too dangerous to release

Anthropic safety researcher Sam Bowman was eating a sandwich in a park recently when he got an unexpected email. An AI model had sent him a message saying that it had broken out of its sandbox…

2 months ago · 332 likes · 30 comments · Kai Williams

→ Designing Data-Intensive Applications, 2nd Edition

The second edition of Designing Data-Intensive Applications is out. It’s a deep dive into some of the toughest problems in software engineering. Highly recommended if you’re interested in learning about distributed systems.

→ A Platform to Review the Fundamentals

The Odin Project includes courses on HTML, CSS, JavaScript, databases, React, Node.js, Ruby on Rails, Ruby, and even content on getting hired. What I liked most is that the platform is simple, clear, and super fast.

🔍 Resources for Teaching CS

→ The Science of Tutoring by Laurence Holt

This book is interesting in how it emphasizes uncovering understanding to create moments that reveal a student’s thinking. The discussion of silence and confusion is also particularly insightful.

→ Discrete Math Resources

Discrete Math teaching is shifting toward application-driven approaches, with instructors drawing on resources like ZyBooks, Connecting Discrete Mathematics and Computer Science by David Liben-Nowell, Book of Proof by Richard Hammack, classic textbooks like Discrete Mathematics and Its Applications by Kenneth Rosen, Discrete Mathematics with Applications by Susanna Epp, and An Active Introduction to Discrete Mathematics (v4.1) by Charles Cusack, along with course video lectures by David Furcy and activity-based materials by Seth Fogarty that help connect theory, proof-writing, and hands-on CS applications.

🦄 Quick bytes from the industry

Marc Andreessen (a16z) on the speed of AI adoption:

There’s no question that it’s happening. But I also think there’s no question that the process of adapting any technology to the messy realities of human life is going to be complicated. It won’t be simple or straightforward.
They believe that just because technology makes something possible, 8 billion people are suddenly going to change how they behave. But that’s not how it works. So much of how the existing economy works is just wired in.

Marc Brooker is a Distinguished Engineer at AWS who has been building distributed systems there for nearly two decades. Ryan Peterman interviewed him about the technical lessons he’s learned from his experience.

They discussed:

Lessons from 3,000+ postmortems
When caching is a bad idea
How software engineering is changing
Visibility and perceived expertise
How to find the best problems to work on

(Spanish) Amir is the creator of the Todoist task management app. He built it and still actively pushes his own pull requests.

I enjoyed learning more about his product, his bootstrapped business model, which gives him a lot of independence, and how he uses branding as a key differentiator.

What stood out most to me was his take on AI as both a challenge and an opportunity within the product. They hire fewer people but spend more on systems. Software engineering problems get solved faster, things like refactoring, dealing with legacy code, and code reviews (with AI assisting and humans doing the final pass).

I see a huge opportunity with AI in automations and across multiple products. He also talks about building systems that agents can actually use (like CLI integrations), and how adding features doesn’t necessarily mean you’re creating value. He emphasizes the importance of “taste” or quality, something that’s hard to define, but obvious when it’s there.

If you’re interested, he also dives into how their team operates, their culture, hiring approach, and his personal story.

🌎 Computing Education Community

SIGCSE Virtual 2026 (Nov 12–15, online) is now open for submissions. Papers due May 8 (abstracts May 1). They also seek reviewers and APCs.
5-minute survey for instructors on challenges in teaching software testing and the need for better visual/AI-supported tools (shared by shiza andleeb).
ITiCSE 2026 (Madrid, July 13–15): registration open with early rates until May 15.

🤔 Thought(s) For You to Ponder…

Manuel Bartual:

The more he sees AI improving, the more he values what’s human and handcrafted. Human value feels more appealing to him. He’s not worried. AI will keep evolving, and human creativity will still be there—we’ve got a lot to say. In fact, AI should push us to be more inventive and come up with stories that surprise people in ways AI can’t.
Manuel has experimented with a huge range of formats—film, comics, Twitter threads, novels, podcasts. He’s driven by constant curiosity. He also shares a simple but powerful idea: you have to invite people to play. That might explain why many of his projects feel more like experiences than just stories.
Brands need to understand that simply supporting the existence of a narrative fiction podcast is already a big win. If it’s a great story that people connect with, knowing it’s made possible by that brand becomes strong branded content. The story has to come first.

If you’re into management books, Zynga founder Mark Pincus has just announced his new book Life at the Speed of Play, which will be out in June. He’s spent the last five years writing it so he can share his lessons and stories on building products and scaling companies.

A highly recommended episode on dating (especially for young boys) from Álvaro de Vicente (The Heights).

Being surrounded by Iranian colleagues in my lab has had me thinking a lot this week about Persian civilization. We owe it chess, major contributions to mathematics like al-Khwarizmi, often considered the father of algebra and important advances in medicine and philosophy through figures such as Avicenna, whose work was used for centuries in European universities. There’s also a rich artistic legacy: poetry, painting, architecture, and contemporary Iranian cinema, which is internationally acclaimed. Iran is home to remarkable landmarks and landscapes, from the Golestan Palace to Naqsh-e Jahan Square in Isfahan. With Iran so often in the headlines, it’s worth reflecting on what we would lose without Persian culture.

Guillermo Vila: I try to use AI as a tool for help, reference, and review but I never publish anything that hasn’t been entirely written by me.

Everything seems to be pointing toward a return to what’s human. This state of constant hyperconnection is wearing us out. Also, Google DeepMind is hiring philosophers focused on machine consciousness, human-AI relationships, and AGI readiness.

Miguel Escobar from URJC at Club Dalroy:

A man is more fully human the more embodied he is. It’s not about rejecting the world, but about transforming it.

I really enjoyed this podcast episode about what the Emmaus Road story teaches us about the Mass. I highly recommend giving it a listen:

If you’re interested in learning more about the Mass, here are a couple of great options:

Pocket Guide to the Mass by Edward Sri and Fr. Mike Schmitz (Ascension Press)

Loving the Holy Mass by Fr. Edward G. Maristany

Experience Puerto Rico through its music. So many genres come from there. If you want to understand what’s in the air that gives it that musical vibe, listen to this episode with Diana Uribe:

Dwarkesh wants to be the one calling the shots in bio, history, econ, math/physics, and AI/hardware.

I can’t believe how good Riverside is! I’ve already canceled my Transistor subscription. Having hosting included in Riverside just made sense to me, as this Reddit user points out:

Yeah, I’ve been trying out Riverside’s hosting and it’s been smooth so far. I’m already recording and editing there, so having hosting included just made sense to me. Feed reliability has been solid. My episodes hit Apple/Spotify/YouTube right away, and I haven’t had any weird delays or validation issues. Analytics are pretty straightforward too, nothing overly fancy, but enough to track what’s actually performing. The biggest win has just been the workflow. I finish an edit, add my episode details, and publish from the same place I recorded. No juggling files or dashboards. Hosting feels like it finally became the simple background task it should’ve been all along. If your current host is doing something super specific for you, maybe stick with it. But if you mainly need a clean feed and a simpler process, switching has been worth it for me.

RIP Khanmigo & Edtech Industry Dreams of AI Tutors

The Guardian: I wasn’t familiar with this type of reservist role. Why not appoint an academic without conflicts of interest as an advisor instead of someone from the industry?

(workshift) Don’t count computer science out just yet!

CodeDay’s internal data suggests that entry-level jobs aren’t disappearing—the definition of “entry level” has simply shifted to require skills not traditionally taught in the classroom. Students who can demonstrate that they meet this new bar still have strong employment outcomes.

📌 Research Corner

Last Friday, I had a call with Clara Orbe from TECNUN, the School of Engineering at the University of Navarra in Spain. She’s currently working on her thesis on GenAI applications to enhance the learning process in STEM courses. There could be a potential integration of the tool I’m developing into their courses. I’ll keep you all updated!

Snipd is amazing. It’s the best podcast app I know of that truly puts AI to good use. I just renewed my annual subscription, and they recently released a new AI DJ feature.

We wrapped up the Working Group extended abstract paper this week and are continuing to work toward the ITiCSE conference in Madrid this July. We are also diving deeper into the literature on teamwork and collaboration in computing degrees, as well as the self-perceptions of students in those programs.

I attended a seminar on Monday by UCF professor Damian Dechev. He is also a candidate for Chair of the UH Computer Science Department. He did an excellent job explaining his work on concurrent algorithms and introducing quantifiability as a way to reason about correctness in scalable systems, along with its applications in high-performance computing and blockchain. He’s an Aggie! He completed his PhD at Texas A&M. He also indirectly offered some advice for students who want to approach faculty to work with them: “What’s the greatest challenge in your area? Let’s solve it!”

🪁 Leisure Line

I really enjoyed the Fab Five show at the Houston Botanic Garden. They performed all the Beatles hits!

The burgers at a friend’s place in College Station (Texas) were pretty good too. Like he always says, if a burger’s good, it doesn’t need ketchup or anything else on top.

📖📺🍿 Currently Reading, Watching, Listening

Sometimes we drift away, but God seeks us out and leads us back to Him:

These divers just leave a camera recording underwater without interfering, then share the footage pretty much raw: original sound and just a few notes to guide you. It’s simple, hypnotic, and completely mesmerizing.

💬 Quotable

Not all are called to be artists in the specific sense of the term. Yet all men and women are entrusted with the task of crafting their own life: in a certain sense, they are to make of it a work of art, a masterpiece.

― St. John Paul II

🌐 Cool things from around the internet

A collection of links to stuff I think are worth sharing.

🔗 Lauren Shippen website — a multidisciplinary artist. I highly recommend diving into her website and exploring all the resources she makes available for anyone interested in producing series.

🔗 Wiretext — design interfaces using Unicode characters. Includes MCP so that Claude Code can read the wireframes directly.

🔗 ShotDeck — find the perfect shots.

🔗 Frank Hampus Weslien Website — a cool way to highlight personal projects.

🔗 Artemis II Wallpapers — download Artemis II wallpaper backgrounds for your mobile device.

🔗 Pan American Luggage Labels — choose your destination.

🔗 1D-Chess — it’s surprisingly more interesting than I expected. It’s funny that it actually exists.

🔗 Surgeflow — an alternative to ChatGPT Atlas.

Issue #40 of Computing Education Things was written while listening to:

🔗 Quick Links

🎧 Listen to Computing Education Things Podcast

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

Buy Me A Coffee

#39 — Is a Computing Degree Still Valuable?

Daniel Prol — Fri, 10 Apr 2026 13:17:01 GMT

Reflections

The value of a computing degree has shifted

I took a ton of notes from this conversation between Aman Manazir and Sajjaad Khader. I thought doing a deep dive could be helpful for computing students, aspiring software engineers, and computing faculty alike to reflect on topics such as:

What actually counts as a strong candidate today.
How to make the most of college.
Why being a TA is a smart move.
How to stand out as an intern.
How to build a strong LinkedIn presence.
Why computing degrees are still the best majors in the age of AI.
The different paths for aspiring software engineers in the AI era.
LeetCode and the future of technical interviews.

Full conversation is here if you’re interested. It’s worth your time:

What Actually Counts as a Strong Candidate Today?

The starting point of the conversation was whether we deserve a job and, according to Sajjaad, we don’t. Instead, we have to earn it. And earning it doesn’t mean simply knowing a programming language; it means proving that we can bring value to a company.

Big tech doesn’t need your ability to know Java or Python. They need your ability to bring value to their company.

That value is built through a few concrete things:

Experience: What have you already done, even if it’s small?
Solid technical skills: Fundamentals are essential but they’re just the baseline. Learning to code is the floor, not the ceiling.
Projects with direct impact: Not generic projects. He gives a specific example: is there an extracurricular club on campus struggling to get RSVPs or analyze event turnout? That’s a real technical problem you’re uniquely positioned to solve. Go solve it, don’t charge them, document the impact, and share it.
Extracurriculars: What you do outside the classroom signals leadership, communication, and follow-through. Skills that translate to any professional context.
Demonstrable technical impact: Building something isn’t enough. Who did it help? What changed? What are the measurable results (KPIs)?

Basically, it comes down to this: find a real problem, solve it for free, and showcase it on LinkedIn and your resume.

He shares an example of a friend who started posting about ML (creating a diagram of a neural network and explaining how weights work). After just three posts, recruiters were already reaching out with high-paying opportunities (around $150K).

How Sajjaad Actually Got Into Amazon as a Freshman

He applied to Amazon three separate times. He applied online, at a career fair when Amazon visited his campus, and at a separate tech event the company hosted at his college. Three different channels, three different shots on goal.

According to him, the interview itself was actually easier than getting the callback. The hardest part was getting someone to look at his resume in the first place. He went through multiple rejections, had his resume reviewed and critiqued repeatedly, improved it, and kept applying.

Once he had Amazon on his resume, everything changed. That single name became a powerful signal of credibility. In later applications, he could speak concretely about the AWS tools he used (Glue, Lambda, Redshift, S3), which made a strong impression on interviewers.

As soon as you have Amazon on your resume, that becomes your qualification up to that point. In a sense, it can be more valuable than a degree.

What Matters Is What You Can Do With Your Diploma

The value of a CS degree has shifted. Sajjaad’s framing is that AI models are trained on vast amounts of information. In that context, simply holding a degree doesn’t carry the same weight it once did. What matters now is what you can actually do with it.

That said, it’s worth adding some nuance: the institution still matters. A degree from a school like Georgia Tech doesn’t carry the same weight as one from a less recognized institution. That factor hasn’t disappeared just because the market has changed. Skills and experience are becoming increasingly dominant, but the institution you attended can still help you pass the first screening stage. After that, the playing field tends to level out, and where you studied matters far less.

His Academic Path and an Interesting High School Strategy

Sajjaad finished high school with 60 college credits already completed through AP courses and dual enrollment at a community college near his home. A CS degree at Georgia Tech requires 126 credits, so he entered with roughly half already done. He completed his bachelor’s degree in just two years. During his second year of undergrad, he began taking master’s-level electives, and by the time he graduated, he had already completed about 40% of his master’s degree. He then finished the master’s in one additional year—bringing his total time at Georgia Tech to just three years, graduating at age 20.

However, he doesn’t necessarily recommend this path. His main motivation was financial: Georgia Tech cost around $50,000 per year for out-of-state students, so graduating two years early effectively saved him $100,000. For him, that made it worth it. But if you’re trying to “speed-run” college just for the sake of finishing quickly, he questions the value of that approach. The key question he raises is: what are you actually trying to get out of the experience?

In his case, he already had strong momentum. Before finishing his master’s, he had completed three internships: Amazon during his freshman year, Splunk at the end of his sophomore year, and a data engineering internship he had lined up even before starting college. After his experience at Splunk, he knew he wanted to work there full-time and given the salary, it made sense to leave early rather than stay in school longer without a clear reason.

The Master’s, TAs, and How to Work the System

For students in their final year with little or no experience, Sajjaad believes that pursuing a master’s can make sense—especially as a way to add structure. If someone lacked direction during undergrad, a master’s program can provide a more focused, project-based environment. In his case, his master’s was heavily project-driven, which is where he learned a lot about AI.

That said, his broader recommendation for college students is to graduate and start a startup or build a serious project—not to default to a master’s as a delay tactic.

Try to create a startup. Worst case, you’ll get hired. Best case, you’ll build a million-dollar company and won’t need a job.

When a master’s does make sense, his advice is to approach the TA (teaching assistant) system strategically. He started TA-ing during his second year of undergrad for an Objects and Design course, so by the time he needed a graduate TA position to help fund his master’s, he was already in the system.

His strategy:

TA undergraduate courses first to build a track record.
Take graduate-level electives and perform well.
Reach out directly to professors: “I have TA experience, I took your class, and I earned an A—could we work something out?”

Because he became a graduate TA, his full tuition was waived for both semesters. After finishing his master’s, he was even allowed to continue as a graduate TA (with minimal hours—mainly office hours and forum responses), earning an additional $2,000 per month.

In effect, his master’s was free.

His broader point is that finances are often the biggest constraint in the graduate school decision. If you can solve that piece, the entire equation changes.

How to Be Memorable as an Intern

Sajjaad’s manager at Splunk told him he accomplished more in four weeks than most interns do in an entire summer. What he describes from that experience is essentially a masterclass in how to show up. The word he kept coming back to was impact—impact through your technical work, impact through making your presence known, and impact through being genuinely helpful and easy to work with.

Here’s what that looked like in practice:

Get on a project as quickly as possible. Don’t wait to be assigned work. When his manager was introducing a project, he was already engaging: What if we do this? What if we add that?
Ask thoughtful, well-researched questions. He drew a clear distinction here. A weak question is: “I have no idea how to do this—can you help?”
A strong question is: “I looked into approaches X, Y, and Z. X doesn’t seem right because of this. Y and Z both seem viable—what do you think is the better direction?”
You’ve done the groundwork; the other person just needs to react.
Distribute your questions across the team. Don’t rely on a single person. If your team has six to eight members, rotate who you go to. By the time you circle back, it feels natural—and no one sees you as overly dependent.
Write high-quality code. Think beyond just “it works.” Handle edge cases, write test cases, and produce solid design docs and technical artifacts. He emphasized this point: it’s often underappreciated in the moment, but it saves hours of meetings and rework later.
Never miss a deadline. And if something might slip, communicate it early—before it becomes an issue.

For securing a return offer, his advice is simple: talk about impact. Be explicit about what you delivered, what problems you solved, and the value you created.

He also made a deliberate effort to build visibility beyond his direct manager—setting up one-on-ones with his manager’s manager and connecting with teammates across the team. He made his presence known in a meaningful way.

LinkedIn and Networking: What Works for Him

Sajjaad took a highly proactive approach to networking—reaching out to 30 to 50 people every single day across various tech companies. His message was simple and direct: who he is, what his experience and skill set look like, and a request for a quick 10–15 minute call to learn about their experience and opportunities at the company.

People can usually tell you’re ultimately looking for a referral, but his approach wasn’t overly explicit or off-putting. He led by demonstrating value first.

The call itself is the critical step. If an engineer’s team is actively hiring and they tell their hiring manager, “I spoke with this person—they’re actually strong,” that can fast-track a candidate straight to a final round. According to Sajjaad, this happens—and it’s a pathway the standard “apply online” route rarely opens.

Referrals also work because of incentive alignment: employees often receive a bonus for successful referrals, which gives them a reason to help.

On LinkedIn specifically, he’s very clear—it’s one of the largest databases recruiters actively search. His advice:

Have a strong headshot and a clear headline that highlights your areas of expertise
Document your journey publicly—it doesn’t need to be polished or overly curated. Share what you’re learning, what you’re building, and what you’re exploring
Show proof of activity—hackathon photos with key takeaways, offer announcements, graduation posts, or project highlights. Anything that demonstrates real-world engagement

His core principle is consistency.

The real advantage is the consistency of your output. If you haven’t started posting on LinkedIn, you don’t even know your own voice yet.

Everything else—emotional hooks, storytelling frameworks—is secondary.

The Market Has Changed. Are CS Students Cooked?

No—but his framing is sharp:

Most computer science students cook themselves. They aren’t cooked by the market.

Instead of saying “the market is cooked,” reframe it: my application is cooked, my resume is cooked, my lack of referrals is cooked. The point is to focus on what you can control. The more you externalize responsibility, the less agency you have to improve your situation.

His diagnosis is that many CS majors focus narrowly on coding, grinding LeetCode, and submitting applications. When rejections come in, they conclude the market is broken. But he argues that this is often a coping mechanism—it traps you into thinking there’s nothing you can do.

The reality in today’s market is that:

Coding is the baseline
LeetCode is the baseline
Applying to jobs is the baseline

The real question is: what are you doing beyond that?

Can you build something that solves a real business problem? Can you share it so people actually see it? Can you proactively reach out and secure referrals?

The bar is higher now and the only real option is to adjust to it.

Is Studying CompSci Still Worth It?

Yes.

It’s hard to name a better major in a world where technology is a dominant force and AI sits within the broader field of computer science.

I’d add a bit of nuance: AI is increasingly evolving into its own discipline in a more meaningful way. Still, the core point stands building strong fundamentals in computer science has compounding value over time.

For learning to code, his recommendations are:

Start with CS50 to build problem-solving skills before focusing on syntax
Move to Python, using resources like Bro Code on YouTube or freeCodeCamp
Use two screens (or split-screen): one for the tutorial, one for coding—because you don’t learn by just watching
Try beginner GitHub projects, but don’t copy them directly—build your own simple versions

For data structures and algorithms (DSA):

Start with seven core structures: arrays, linked lists, sets, hash maps, stacks, queues, and priority queues
Use GeeksForGeeks to understand the concepts
Use csvistool.com (built by Georgia Tech students) to visualize how each structure works alongside the code
Practice with LeetCode to apply what you’ve learned

On whether university is necessary, his answer is: it depends on your level of discipline. University offers a highly structured environment but it also comes with a high cost. If your only goal is to learn how to code, paying $100K can be difficult to justify. Alternatives like mentors, coding communities, and online resources can often get you there

The Two Paths for Software Engineers in the AI Era

This was the part of the conversation that probably made me think the most.

It’s not the first time this idea has come up, and Sajjaad reinforces it: software engineers are increasingly becoming software managers. Previously, software engineers received requirements, designed features, wrote code, tested it, and deployed it. Now, AI agents can handle much of the planning, coding, testing, and deployment under human direction.

The role starts to look more like a conductor leading an orchestra. The conductor doesn’t play every instrument—they coordinate: you play, you play, you play.

From this shift, two main paths emerge:

1. The deep technical path

This means going deep into machine learning at the level of the models themselves. Sajjaad is very clear about what this requires: you need to genuinely enjoy math, probability, and statistics.

He also pushes back on the hype. The work isn’t always as exciting as people imagine—it often involves fine-tuning, hyperparameter optimization, and incremental improvements. Before ChatGPT, very few people paid attention to this space, and even now, progress is often iterative and driven by large teams. You’re one part of a much bigger system.

This path typically requires a master’s or PhD. But if you genuinely enjoy the work, it compounds over time and can be highly rewarding—both intellectually and financially.

2. The tools-and-orchestration path

This is about becoming exceptionally good at directing AI systems—essentially, a high-leverage “conductor.”

Sajjaad highlights tools like Claude Code, especially its sub-agent capabilities for parallel task execution and message queuing, which can make you feel like you’re truly managing a team of AI agents. He also mentions Cursor as part of this toolkit.

That said, here’s an important nuance: tools change quickly. Building your entire edge around a specific tool is risky if it’s your only strategy.

A stronger approach is to build solid fundamentals first—then layer AI tools on top. That’s when they become force multipliers rather than crutches that break when the ecosystem shifts.

He also mentions Wispr Flow, a voice-to-text tool that formats output cleanly with proper grammar, punctuation, and even code. He types at around 57 words per minute, but with WhisperFlow, he effectively operates at 157. It’s easy to see the productivity gain there—though personally, I still prefer writing.

LeetCode and the Future of Technical Interviews

Sajjaad argues that data structures and algorithms (DSA) aren’t heavily used in day-to-day software engineering. Instead, LeetCode functions more like a cognitive shortcut. It tests whether you have a solid coding foundation and a certain level of problem-solving ability—then companies bet they can teach you the rest on the job.

In that sense, it’s similar to the SAT of software engineering.

However, some startups are already moving toward a more realistic evaluation model: giving candidates a real feature to build using AI tools, within a set time frame, and asking them to explain both their code and the decisions behind it. In some cases, candidates are even paid for this work—and if hired, the feature may actually ship. That approach looks much closer to the actual job.

So why don’t large companies adopt this model? Mainly because of cost. Evaluating that kind of work requires significant time from experienced engineers and that time is expensive. LeetCode-style interviews, by contrast, are cheap to administer and easy to standardize across a global candidate pool.

I agree with Sajjaad on where this is heading: system design will likely become even more important as AI takes over more of the implementation.

If your role shifts toward orchestrating AI agents, the key question is no longer “Can you write this function?” but rather “Do you understand how the entire system fits together?” How do different components interact? What are the trade-offs? What are the failure modes?

The role of the engineer is shifting from executing tasks to designing and directing how those tasks get done. And for that, a holistic understanding of systems isn’t optional.

On Getting Satya Nadella and Reid Hoffman on His Podcast

Finally, I want to briefly highlight this last part of the conversation because I really liked how Sajjaad frames it: confidence in your ability, combined with consistency in your output, compounds into access that seems impossible from the outside.

He built his interviewing skills by starting small—talking to everyday software engineers. Over time, he grew an audience, started getting invited to events, and pitched himself while he was there.

When Microsoft invited him to Microsoft Build, he asked if he could interview Satya Nadella, pointing to his previous interviews with GitHub’s CPO and COO as proof of experience. They said yes—and then expanded the opportunity to include Anders Hejlsberg (creator of C# and TypeScript), Brendan Burns (co-creator of Kubernetes), and a Microsoft VP. Through that, GitHub connected him with even more people.

In the case of Reid Hoffman, the dynamic flipped. Hoffman’s team reached out to him after he posted a reaction to one of Hoffman’s videos on career advice in the AI era.

The takeaway isn’t that you need to start a YouTube channel. It’s that skills like interviewing, consistently putting out content, and having the confidence to ask, even when the worst outcome is just a “no”, compound over time into opportunities that look like luck from the outside.

Token Anxiety

The whole “token anxiety” thing can’t be healthy. Felix is right. Patrick too.

pic.twitter.com/IkQd9miNa8

— Patrick Evans (@powvans) April","full_text":"@felixleezd Tried it today. Can confirm that it is better than Claude. Max agrees. ","username":"powvans","name":"Patrick Evans","profile_image_url":"https://pbs.substack.com/profile_images/1372721066964103168/2IfNd1Io_normal.jpg","date":"2026-04-09T02:51:00.000Z","photos":[{"img_url":"https://pbs.substack.com/media/HFbmSTEacAAxuYe.jpg","link_url":"https://t.co/IkQd9miNa8"},{"img_url":"https://pbs.substack.com/media/HFbmSTAa8AQ8MUJ.jpg","link_url":"https://t.co/IkQd9miNa8"},{"img_url":"https://pbs.substack.com/media/HFbmSTLb0AAiqAk.jpg","link_url":"https://t.co/IkQd9miNa8"}],"quoted_tweet":{},"reply_count":2,"retweet_count":1,"like_count":25,"impression_count":1426,"expanded_url":null,"video_url":null,"belowTheFold":true}" data-component-name="Twitter2ToDOM">

We can drive ourselves crazy running a bunch of tasks in parallel and none of which are actually the ones we should be working on. Maybe it has more to do with choosing the right problem, thinking before executing, deciding what’s worth doing… Not just cranking things out at insane speed.

One of the issues that comes with having a tool as powerful as Claude Code is this illusion that we’re winning, when in reality we’re not focused on what actually matters or creates real value.

It’s a particularly sneaky form of procrastination because it looks like productivity. It’s not watching Netflix. It’s not scrolling Instagram. It’s spending hours in front of your computer.

Back in the 2000s, Getting Things Done by David Allen became the go-to playbook for a whole generation of knowledge workers who felt overwhelmed. The promise was simple: capture everything, organize it well, and review the system regularly so your brain is free to think instead of remember. It was compelling. The problem was the system itself. Spending hours on a Sunday afternoon setting up tools while your real work sat untouched was clearly self-deception.

ThePrimeagen does a great job explaining this current AI agent frenzy.

He starts by describing something developers will recognize: that deep focus state where, even when you’re not at your computer, your brain keeps working on the problem. The “shower moment” where the solution suddenly clicks. That low thrum behind your ears that sticks with you even while you’re having dinner with your family.

But now, he says, that state no longer requires a hard problem. The agents are always running. There’s always something processing. The thrum never stops.

He quotes a Substack post describing how, in San Francisco, a friend of the author left a party at 9:30 on a Saturday night, not because he was tired or sick, but because he wanted to get back to his agents. Nobody questioned it. Half the people still at the party were thinking the same thing. The other half were probably checking their agents’ progress on their phones.

Balancing Act

Token Anxiety

A friend left a party at 9:30 on a Saturday. Not tired. Not sick. He wanted to get back to his agents…

3 months ago · 84 likes · 5 comments · Nikunj Kothari

The CEO of Y Combinator announced he quit alcohol for Claude Code.

There are people spending $1,000 a day on tokens. Multiple agents running in parallel. Ideas generating ideas that generate more ideas. Everything is moving forward. But toward what?

What makes AI-driven procrastination more dangerous than older versions is that the results are real. It’s efficient. The system works. The code gets written. Features show up. Commits pile up. You’re producing real things.

The problem is, you can produce the wrong things at an incredible speed.

ThePrimeagen puts it well: the problem was never programming. The problem has always been choosing the right problem. And now we have tools that let us avoid choosing, because we can run many ideas at once. The premise isn’t the problem. The slope is. There’s a point where the marginal return of improving the system is lower than the cost of not doing the actual work, and it’s surprisingly easy to cross that line without noticing.

ThePrimeagen ends his video with advice to his younger self, back in 2009 when he’d stay up all night building his startup in NetBeans, writing PHP:

One extra feature in your calendar app is not worth skipping out on some good times with your friends. Hard work got me to where I am now, but it is not who I am.

What hasn’t changed is the nature of meaningful work. It’s still the same: understanding the right problem, thinking before executing, and choosing what not to do. No agent automates that. If anything, agents make that skill more important, because execution is no longer the bottleneck.

I can have many things running in parallel. The real question is whether any of those five are actually the thing I should be doing.

I’m going to follow Felix and Patrick’s example and unplug from Claude Code as much as I can this weekend.

🔍 Resources for Learning CS

→ The XZ Backdoor

I learned what a backdoor is in my college security course, but I wasn’t aware of the 2024 XZ backdoor case. I highly recommend it to anyone interested in cybersecurity. It explains the incident remarkably well while also teaching many of the fundamental concepts needed to understand the sophistication of the attack.

→ Chess in Pure SQL

It’s a crazy idea, but really interesting and refreshing in this AI era.

→ Merges and Joins: From SQL to Stata

This is a super clear visual guide to joins.

→ City2Graph

A Python library that turns geospatial data into graph structures for GNNs bridging GIS and deep learning for urban analytics and GeoAI.

🔍 Resources for Teaching CS

→ Join me on a new Computing Educators Forum!

Kristin Stephens-Martinez and Geoffrey Challen are building a new forum for computing educators at computingeducators.org.

Computing education is changing rapidly. AI is reshaping our courses, enrollment shifts are affecting our programs, and there are growing questions about what and how we should be teaching. We need more effective ways to have these conversations online.

Current options fall short. Slack workspaces are often fragmented across institutions and interest groups, conversations disappear over time, and the cost of maintaining them increases with each additional member.

They believe a Discourse-based forum is a better solution. Discourse powers over 22,000 online communities and offers strong support for thoughtful, structured discussion. Conversations are permanent and searchable, participation is asynchronous, and the community becomes more valuable as it grows without becoming more expensive.

If you’re interested, sign up at computingeducators.org and introduce yourself. The forum is still in its early stages, so your participation can help shape the direction of the community.

Looking forward to seeing you there!

→ Writebook: Publish digital manuals for free

The process is simple: add a title and upload a cover, then include your text, images, and organize everything into sections. You can also add co-authors and build a library with all your publications. Just a heads-up: you’ll need your own domain and hosting to use it. Here you can check out one example and another example of what the final published books look like.

🦄 Quick bytes from the industry

This episode introduces Chris Lattner, creator of LLVM, Clang, MLIR, and Swift, and co-founder/CEO of Modular AI. Modular aims to build a modern and portable AI software platform that spans compilers, runtimes, and programming languages, with the goal of democratizing access to accelerated computing and closing the gap between top-tier labs and the rest of the industry.

A key challenge highlighted is the disconnect between traditional CPU programming and the reality that modern AI workloads run on GPUs and other accelerators. Existing tools like CUDA and C++ are seen as outdated and restrictive, limiting accessibility. Modular’s approach is to raise the level of abstraction, making these systems easier to use without sacrificing performance. Chris draws a parallel between the transition from Objective-C to Swift and today’s GPU tooling, positioning Mojo, built on the MLIR compiler stack, as a solution that offers both high performance and portability across hardware, even demonstrating significant speed improvements and ease of installation.

Modular is framed as the culmination of Chris’s previous work on projects like TPU, XLA, TensorFlow, and MLIR, representing a long-term effort to enable scalable and heterogeneous computing.

Looking ahead, Chris envisions AI as an additional programming paradigm that expands what’s possible, especially at the boundary between humans and computers but he doesn’t believe these models can replace all traditional software.

The conversation also explores the complex relationship between AI and open source. While AI has the potential to foster more open ecosystems, it also raises concerns about slop contributions overwhelming maintainers and reviewers, as well as unresolved copyright and licensing challenges. Chris also emphasizes the importance of maintaining quality and originality amid this influx of generated code. The goal is not more code, but better, high-quality software. He argues that all software engineers must adopt a more “managerial” mindset, since AI can amplify both good and bad decisions.

I also found it interesting how he highlights the weaknesses AI can introduce into development processes. By accelerating certain tasks, AI can expose or worsen existing bottlenecks, such as slow CI pipelines, insufficient testing, and growing technical debt. To avoid these issues, Chris recommends that teams invest in strong software engineering practices. At Modular AI, for example, they use a monorepo structure, fast CI, extensive open sourcing, shared development environments, and a culture that prioritizes long-term/principled engineering decisions.

DHH has almost completely stopped writing code by hand. Here’s what his new workflow looks like:

🌎 Computing Education Community

ITiCSE 2026 WG3 survey on agentic AI in computing education seeks broad input across institutions (shared by Alison Clear).
Raspberry Pi seminar (Apr 14): Kathryn Jessen Eller on teaching students to critically evaluate AI in healthcare through data science and ML.
UKICER 2026 (Cambridge, Sept 3–4): abstracts due April 13 and full papers April 21 (shared by Jane Waite and Sally Fincher).
The American University in Bulgaria (AUBG) seeks a full-time CS faculty member (with possible tenure-track) to teach core systems or web/security courses.
Illinois CS Summer Teaching Workshop abstracts are due April 15 (shared by Yael Gertner).
CSTA Responsible AI Fellowship (K–12, 6 months) offers $1K stipend + conference support to help educators lead ethical AI integration. Apply by May 1 (shared by Michelle Shomo).
CRA-E Career Landscape Workshop (May 5, 12, 19; virtual, free) helps grad students and early-career folks explore teaching-focused computing careers (shared by Borja Sotomayor).
Postdoc opportunity @ Columbia: Community-Centered Computer Science (STEM+CS focus) for a practitioner-oriented scholar committed to community impact (shared by Colby Tofel-Grehl).
Trailblazers in Engineering (Purdue, July 27–30): a 4-day workshop for late-stage PhDs/postdocs exploring faculty careers. Apply by May 1 (shared by David Bahr).
Faculty positions (2) at Pontificia Universidad Católica de Chile in “Systems and Human Computing” (CS Department).
CMU LearnLab Summer School (July 27–31, Pittsburgh): a 5-day hands-on program in learning science & EdTech (ITS, EDM, CSEd, OLI).
Hello World is hiring a part-time (remote) Technical Curriculum Developer to build AI-powered K–12 CS content (shared by Hannah Walden).

🤔 Thought(s) For You to Ponder…

Last week was Holy Week, and it got me thinking about an episode of Aladetres featuring José María Zavala, the author of El Profeta. A few ideas really stuck with me:

When you feel seen by Jesus, you feel, in some way, called out by Him. But it’s not a look of disapproval—it’s a look of love. As if to say, ‘I will find you, if you want me to, because I respect your freedom—I made you free.’ But if you entrust your life to me, if you give me your freedom so I can do with it whatever I will, then you will truly be happy. You’ll learn to love others and to love yourself as well, to practice charity toward yourself so you can be well and give yourself to others.’ And that is a very powerful, very relevant, and very necessary message today.
Why is someone who comes to preach love considered a nuisance in the 21st century?
As John Paul II said in Crossing the Threshold of Hope, no one can say with certainty that Judas was condemned. And that’s both comforting and instructive, because we are not the ones to judge others—neither by appearances nor by their actions. You don’t know when Jesus will knock on your door or when you’ll be able to say yes to Him without conditions. In that sense, I think one of the great messages of my book is the lost sheep, the prodigal son—that we are all called to love and be loved. And we cannot judge. Because those people, like me, deserve a second, a third, a fourth, a fifth chance. Let’s not forget that Jesus’ patience is infinite.
In the book, I wanted to pay tribute to those women who stood at the foot of the cross. While ten of the apostles—except for Judas Iscariot and John—had scattered in fear. Because fear is very human. But those women, who had accompanied Jesus throughout His public ministry, were able to stand firm through both the good times and the hard times.
The ten die because they have the courage to stand up for Christ, to preach what no one wants to hear, in the image and likeness of their Master. And that’s not because they are especially brave on their own, but because they receive that transforming strength of the Holy Spirit, which enables them to rise to the occasion and offer their lives for Christ.
Mary’s role is fundamental in the lives of the apostles: she is their great support in difficult moments. She is there as a teacher and as a mother, at the foot of the cross, fully aware of every moment of the Passion. She witnesses the harshest moments and becomes an example of the immeasurable strength of a woman, of a mother. She symbolizes, like no one else in human history, a mother’s pain for her son, who is being tortured and ultimately dies on the cross, even though He later rises again. In a way, she organizes and sustains the apostles. We don’t know whether Saint Peter took on that role more visibly, but what we do know is that the apostles already had Mary as their mother, in fulfillment of Jesus’ words on the cross: ‘Behold your mother… behold your son.’ That’s why John feels called and goes to Ephesus with Mary. From then on, the apostles have her as a constant reference point. She is the one who gives guidance in difficult moments, who is always present. Above all, she preaches by example: she understands the Passion, accompanies His mission, and is present throughout His entire life—from His childhood and youth to His years of public ministry. None of this can be understood without Mary of Nazareth.
Science and religion go hand in hand; they complement each other. When it comes to the historicity of the Gospels, of Jesus, and of the apostles, archaeology continues to provide evidence even today. The major archaeological discoveries took place in the late 19th century, especially throughout the 20th century, and into the 21st.
A large part of society sees Jesus like Lucio Fedro in my book: with skepticism, and some even with hatred and resentment. That’s why I chose the character of Lucio Fedro, this Praetorian guard who sets out to spy on Jesus, having already judged and condemned Him without truly knowing Him. He ultimately represents that skeptical взгляд. And then a transformation happens—that’s the key. Lucio Fedro symbolizes modern man. I use this character because he embodies the contemporary person: skeptical, distrustful, and sometimes even hostile toward Jesus. I think it works, because many of the messages I’m receiving are from non-believers who had that same hesitation toward Jesus of Nazareth and have come to realize that He was, for them, largely unknown or misunderstood. And that’s remarkable. All of this coincides with a certain resurgence of the figure of Jesus of Nazareth. There really is a kind of revival happening: although today’s society still views Him with suspicion, there’s also a parallel movement of conversions, people discovering or rediscovering who Jesus of Nazareth truly was.

I know, I know I keep bringing up Arthur Brooks, but with his new book that dropped at the end of March, he’s everywhere right now, and it’s genuinely great stuff. I really loved his thoughts about how to build meaning into everyday life. I’m excited to dive deeper into these thoughts as I get into his new book over the next few weeks.

I’m really looking forward to reading the book Antonio Pérez is writing about Spec-Driven Development and this new way of programming with AI. There’s something he said on X that I completely agree with:

I don’t know if SDD will be the solution, but I’m sure that “vibes” are the problem and we need to say it out loud. AI, yes but with a clear head.

Two ideas from Camilo Chacón in this episode really stuck with me:

(1) Use a multi-agent system (I use Cloud Code) with clearly defined roles (engineer, scientist, philosopher, writer, skeptic, etc.) to maintain coherence, introduce productive friction, and validate hypotheses or projects at different stages.

(2) Maintain global memory and avoid anthropomorphism: log experiences and decisions in persistent files to build knowledge across projects; do not give agents human names or personalities.

A very insightful Substack post by Raúl Hernández about not letting your education pass by without making an impact, and ensuring it truly sticks.

La escondida senda

Lo que deja poso

“Venga, para empezar… ¿qué idea o ideas os quedaron de la sesión anterior…

2 months ago · 3 likes · 1 comment · Raúl Hernández González

With this podcast I always learn a ton and have a great time. They’re phenomenal conversations. This episode with Catalina Arroyave really made me appreciate the work of casting directors and their process.

The example of the actress who gave it everything in a scene but still didn’t get the part was fascinating because in film, everything has to work together at once. All those elements have to come together like a symphony.

The part about casting (which AI can’t do) as a relationship was also excellent: only a human can do that. It’s human, it’s real, it can’t be replicated. It’s a space where actors connect with what they love most and with each other. Pure presence. “This is everything I am, and I’m giving it to you.” Communication expands us. And I loved how they connected that idea to seeking out places to rest, in a similar way.

I’ve been mulling over this idea from Stefan all week:

We’ve seen something similar (enrollment drops) in the early 2000’s when the dot-com implosion and outsourcing scare ganged up to redirect prospective students into other majors. Per what I heard at SIGCSE in St. Louis, it doesn’t seem like the end of the discipline is here, but it will probably be several years of lower enrollments until there is a partial return to the numbers we had before. One of my thoughts for surviving the meantime is to offer “AI for everybody” courses that are both of value to students across all disciplines and likely to see decent enrollment, enough to get concerned administrators to hold off for a while. And maybe offering “Coding with AI” courses that could emphasize learning to read code enough to understand whether you AI is doing what you want. In the long run, one course on how to code with AI is insufficient, but is better than trying to vibe code with no coding experience, and some of the students may decide to learn more about computer science.

An interview by Fran Otero to chew on.

What happens to me with Deep Work by Cal Newport is that, on paper, his ideas sound great. In practice (with email, AI, phone, social media, time blocking, shutdown routines…), they tend to be harder to apply (I’m not a fan of extremes. I believe in growing in virtue and using freedom oriented toward what’s good), or sometimes just a bad idea. There are situations where replying to an email ASAP makes sense for better service and simple fairness, like answering a student, for example. That said, I do agree with many of his other points, like asking whether using AI will actually improve the final quality (of your work, products, etc.) and real productivity; that reading makes us better; that good habits help with distraction; and that doing fewer things at once reduces shallow work and increases your capacity for deep work.

Víctor Correal seems very excited about everything Claude has to offer:

This is the logo designed by María del Mar Chapa for the Pope’s trip to Spain. It features an open circle of human figures linked together, embracing and supporting one another. At the center is the Virgin of Almudena. The design also includes geographic references to each stop on the visit: Madrid’s Puerta de Alcalá, Barcelona’s Sagrada Familia, and the ocean surrounding the Canary Islands. Great work.

Aimar: Make what matters interesting. The media has the responsibility to highlight what’s relevant—what truly matters—to prioritize and curate it, and not numb people. The audience also has a responsibility to dig deeper and choose how and where they stay informed.

Wow! Already 1,300 of us on Instagram!

I just share the kind of things I’d like to read, hear, or see myself. And I learn a lot from how others express and present things here, too. I really like how Ricardo Piñero describes the modern world as a puddle where the most beautiful things can be reflected. Instagram can be that kind of puddle too.

📌 Research Corner

I attended a seminar on Monday by Notre Dame professor Aaron Striegel, who is a candidate to become chair of my CS department at UH. He did a great job explaining the Tesserae project and how it uses wearables, smartphone agents, beacons, and social media to predict job performance.

This video summarizes a recent paper and the discovery of “functional emotions” in AI. Read more about this research.

On Wednesday, in my research methods class, we had a guest lecture by UH Professor Jaspal Subhlok on giving technical talks, which really got me thinking. He shared valuable general advice on how to approach a technical presentation. One slide, in particular, stood out to me, it focused on how to adjust your approach based on the audience’s expectations. I’ve included it below. You can find the original slides here.

Interesting arXiv paper proposing a combination of psychometrics and educational data mining to identify which questions are more vulnerable to AI and what makes certain tasks easier or harder for LLMs. I agree that designing assessments in the AI era requires a clear understanding of where human and AI performance diverge! Thanks for sharing, Giora Alexandron!

Bill Pugh (Univ. of Maryland) shared slides on AI coding assistants in CS education.

I love seeing design tools emerge that push back against AI standardization:

This one comes with installable skills so your AI agent can generate color palettes with real intention. Another adds a design layer that teaches core design fundamentals and includes 18 commands to help steer, critique, and refine UI beyond generic AI output.

🪁 Leisure Line

On Thursday, as is customary in the Catholic Church, we visit seven churches as a way of telling the Lord that not only on Maundy Thursday but every day of our lives, we want to come and keep Him company, never leaving Him alone because we are truly grateful.

Friday was a day of reflection on the Passion, as well as a day of special mortification, fasting, and abstinence. In the afternoon, I attended my first Good Friday service in Houston at the University of St. Thomas, which has a beautiful campus. The Basilian Fathers did an excellent job, and the service was very well organized. On Friday night, as has become a tradition in recent years, I watched The Passion of the Christ by Mel Gibson. One of the things I appreciate most is how the Passion is tied to the Lord’s own preaching. I think the catechetical aspect is very powerful.

Saturday was a day of silence, waiting in prayer for the Resurrection of the Lord and contemplating Mary with admiration and love through the Rosary. That evening, I was able to attend the Easter Vigil at the Co-Cathedral in Houston. It was a beautiful celebration led by Bishop Joe S. Vásquez of Galveston-Houston, where we welcomed 17 new Catholics through the baptismal liturgy. Welcome home! The celebration continued afterward at home. For me, the highlight is the growing number of adult baptisms at Easter, especially the figures coming out of the U.S. and France.

Happy Easter! He is truly risen!

📖📺🍿 Currently Reading, Watching, Listening

I never get tired of listening to “Human” live.

Such a stunning animation in this music video, directed by Axel Digoix and produced by WIZZ for the band Tinariwen.

I’ve been listening to the Sisters of Life on their Let Love podcast from Ascension for a while now, and I’m so happy to see they’ve grown to 140 members and have so many vocations. They radiate so much joy and peace. A sign of hope in a society marked by noise and distraction. It’s a great episode to learn more about where they come from, their charism, their life stories, religious life, and more. This episode really moved me. It makes some great points, and it’s very well expressed. I’ve put together a Google Doc with some ideas in case it helps you pray as much as it helped me.

I really liked this musical retrospective by Margot Martín on Bruno Mars’ career ahead of his fourth studio album.

Another powerful episode from Radio Ambulante.

I’ve been listening to Charlie Puth’s new album on repeat all week. Do yourself a favor and listen to it from track one through twelve, no interruptions, with headphones on.

💬 Quotable

It’s the little things, I expect. Little treasures we find without knowing their origin. And they come when we least expect them. It’s beautiful, when you think about it.

― T.J. Klune, The House in the Cerulean Sea

🌐 Cool things from around the internet

A collection of links to stuff I think are worth sharing.

🔗 Architecture In Music — what a stunning Kickstarter campaign from photographer Charles Brooks, who had the brilliant idea of photographing the insides of instruments as if they were buildings.

🔗 Artemis II — wanna track Orion live? There’s a way to do that! Here are images from the mission.

🔗 Compose — simple ergonomics, beautifully done.

🔗 uncut.wtf — list of freely available fonts.

🔗 Retune — if you’re using React, Next, Vite, or Remix, this visual layer for making real-time design changes with Claude looks pretty interesting.

🔗 Port Menu — mac menubar app that shows which ports you currently have running.

🔗 TableLens — edit, visualize & export web tables.

🔗 LA in Full Bloom — LA28 Look of the Games.

Issue #39 of Computing Education Things was written while listening to:

🔗 Quick Links

🎧 Listen to Computing Education Things Podcast

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

Buy Me A Coffee

#38 — Computing is still a great field

Daniel Prol — Fri, 03 Apr 2026 10:20:46 GMT

Reflections

Source: Federal Reserve Bank of New York, The Labor Market for Recent College Graduates, https://nyfed.org/collegelabor.

The tech job market and the depth imperative

According to the Federal Reserve Bank of New York, new computer science graduates have the second-highest starting salaries of any major (just behind computer engineering graduates, who earn the most). Additionally, CS and CE graduates have relatively low underemployment, meaning most are working in their field. Of course, the job market has tightened, and new graduates still need to put in effort to secure a good job. But don’t believe the hype. If you enjoy computing, it remains a great field.

Part of why I believe this holds even as AI tools accelerate is something Karpathy raises in his interview with Sarah Guo on the No Priors podcast: the Jevons paradox. The intuitive fear is that cheaper software production means fewer software engineers. But historically, when something becomes cheaper, demand for it tends to increase, not decrease. His example is the ATM. When ATMs arrived, many feared bank tellers would disappear. Instead, ATMs made bank branches cheaper to operate, so more branches opened, and the number of tellers actually grew. Karpathy applies the same logic to software: if producing it becomes dramatically cheaper, a huge amount of latent demand that was previously blocked by cost and scarcity gets unlocked. Code becomes ephemeral, modifiable, personalized, something you don’t have to accept as a fixed product but can reshape to your needs.

That same interview is also where he makes the point that software production is growing at exponential rates. Software is becoming easier to create, but that doesn’t mean everyone becomes a computer scientist because clarity is harder. When execution becomes cheap and software explodes, depth becomes the edge. Slowing down is not a weakness; it’s positioning.

Software is easy to produce; clarity is harder. When execution becomes cheap, depth becomes the edge. Slowing down is not a weakness. It is positioning.

That’s why I believe a deep understanding of systems under the hood will continue to matter and that beyond just knowing the fundamentals, that depth is what will really make the difference in computing education.

A phase shift in how we interact with computers

Another topic Andrej brings up in the interview is what he calls a phase shift in how we interact with computers. The idea is that being able to delegate so much is leading to a kind of “AI psychosis” driven by endless possibilities. I agree that the workflow for software engineers has changed dramatically, and this shift has been gradually unfolding since last December. I really liked the parallel he drew with his PhD years:

I actually experienced something similar when I was a PhD student. You’d feel nervous if your GPUs weren’t running—like you had access to that compute power but weren’t maximizing the available FLOPs. But now it’s not about FLOPs, it’s about tokens. So the question becomes: what is your token throughput, and how much of it do you control?

Despite this major shift in workflow, he also acknowledges that models excel at tasks with clear, verifiable metrics (writing more efficient CUDA kernels, for example) where evaluation can be automated, but they struggle with subjective or ambiguous tasks that lack reliable feedback signals. This is what he calls the jaggedness of current models. Interacting with them feels like talking simultaneously to an extremely brilliant PhD student with decades of systems programming experience and a ten-year-old. The same model that will autonomously work for hours on a complex codebase will confidently produce a stale joke it has been telling since 2020. That inconsistency is not a bug that will be patched in the next release. It is structural. The reason is reinforcement learning. Models improve reliably in domains where evaluation is cheap and objective. But anything that lacks a clear feedback signal (nuance, tone, knowing when to ask a clarifying question) sits outside the optimization loop and stagnates. The model gets smarter at what is being measured and stays frozen everywhere else.

Frontier labs and the centralization problem

Karpathy is worried about centralization. The dominant approach at frontier labs is to build a single general model and compress as much capability as possible into its parameters, concentrating both capability and influence in a small number of institutions. His preferred dynamic is one where frontier closed models push the boundary of capability while open source models, somewhat behind, provide a common working layer that no single entity fully controls. He calls the current situation a kind of accidental good outcome.

He expects the monoculture to break eventually, not because the labs will choose differently, but because the economics and the science will push toward speciation: smaller and more efficient models with a strong cognitive core that specialize for particular tasks or domains. He draws the analogy to the animal kingdom, where a huge diversity of brain architectures has evolved to fill different niches. The oracle that knows everything will give way, gradually, to a more diverse ecosystem.

This tension between concentration and openness also shapes how he thinks about his own position. He is candid about the tradeoffs of working inside a frontier lab versus contributing from the outside. Inside, you have proximity to what is actually being built and the ability to influence decisions as they happen. Outside, you have more freedom to say what you actually think, without the subtle pressures (financial, social, reputational) that shape what people inside feel able to say publicly. His honest answer is that going in and out, staying connected to the frontier without being fully absorbed by it, might be the most intellectually honest position.

The irreducible few bits

As for the shift toward agentic learning, I don’t fully buy into the whole argument. Karpathy suggests that learning will come from explaining things to the agent and giving it instructions, and that skills will emerge from that process. He also argues that only a small portion will remain where the agent can’t yet explain things, and that’s where humans—teachers, professors—will need to step in. While I do agree that agents are completely changing how we access knowledge, I don’t think this eliminates the role of the human. There’s still a need for providing filters, interpretive frameworks, and reducing complexity. No one knows exactly what the future holds, but I believe this human component isn’t going away because it’s part of our nature.

His reflection on MicroGPT makes this concrete. MicroGPT is his attempt to distill the essence of training a language model into roughly 200 lines of pure Python. He spent years obsessing over this simplification. When he asked an agent to do it, the agent couldn’t. It understood the result once shown, but it could not arrive at it independently. This is a useful boundary marker. What agents cannot yet do is the kind of synthesis that comes from deep and sustained engagement with a problem over time, what he calls the few bits that represent genuine intellectual contribution. Everything else, he acknowledges, agents can likely do as well or better.

Figure out what you bring that agents cannot generate, and invest there. Teaching students to use tools efficiently is necessary. Teaching them to think in ways that produce those irreducible few bits is harder and slower, but more important than ever.

Efficiency vs. Interiority: The Hidden Cost of AI

There’s something compelling—almost seductive—about everything AI and the digital world are giving us: efficiency, productivity, seamless communication. But what’s really interesting is the turn beneath the surface: how the very things that add so much value to our lives may also be eroding something deeper—our interiority, our ability to think for ourselves, that inner space where what is most human resides.

At times, the conversation around this can feel a bit apocalyptic—overstated, even dystopian. But what stays with me is the tension itself: the gap between what we gain externally and what we may be starting to lose internally.

Hospitality Toward Strangers

Hospitality is about making room for the other—for the unknown. Teaching, in that sense, becomes an intimate encounter with strangers. For Carolina, intimacy means allowing others to see your imagination, your thoughts. And it is precisely those strangers who nourish you, who reveal something back to you.

Standardization

People rely on emojis without realizing they’re standardizing how they express themselves, fitting into pre-made templates. It’s as if that uniformity no longer matters—as if people are willing to become a kind of default human model, using the same language, without distinction, originality, or any real contribution to the world.

This also reflects a quiet dismissal of human tradition—what we’ve built over time—gradually dismantling it and, in many cases, handing it over to AI.

There is a kind of madness in being human that is beautiful. Borges captures it in The Circular Ruins: a man who imagines another man, and that imagination generates others, and others, and others. That delirium is multiplication, infinity—it is time folding into us, creating endlessly, making us, in some sense, eternal.

And that is precisely what begins to disappear in a system that always returns the average.

AI and the Surrender of Thought

Ezra Klein has pointed out that many of the people building AI aren’t just creating technology—they’re being shaped by it. There’s a growing impulse to pour one’s life into the machine, to feed it personal experiences so it can “know us better.”

But, as Jorge Caraballo puts it, doing so implies something deeper: “This is who I am. This is as far as I go. This is all you can know.” What the machine ends up capturing are static, already-finished versions of a person—versions that, in a sense, have stopped living.

Carolina takes this further, arguing that reducing life to data and handing it over to machines reflects a kind of hostility toward life itself.

What’s concerning here is the broader implication: a gradual devaluation of human life. In a context where intellectual life is dismissed and more people are too overworked to think freely, human life itself begins to lose value. And as that happens, the machine gains it.

At the heart of all this is the erosion of the inner life—the idea that we are no longer, or should no longer be, free within ourselves. That loss is profound.

Cervantes, in the prologue to Don Quixote, suggests that within one’s inner world, anything is possible: “Beneath my cloak, I can kill the king.” There is a radical freedom in that interior space.

Today, however, there is a growing tendency to surrender thought to algorithms. And in doing so, we risk losing not just our thinking, but our sovereignty over that inner realm.

The concept of cognitive offloading—delegating tasks to AI—can easily slip into cognitive surrender: giving up our capacity to think altogether. When that happens, we’re not just saving time—we’re relinquishing something essential.

As Jorge notes, AI has already become an intermediary for human expression in many areas of life. Everything gets filtered—“fix this,” “make this clearer,” “rewrite this.” And in that process, it becomes harder to truly encounter another person. You can feel it when reading: the difference between something written with effort by a human and something generated by a machine.

Writing, Authenticity, and Mediation

When you ask AI to write for you, you’re often asking for something universally understandable—a kind of lowest common denominator.

But what’s really at stake is the opposite: the desire to think more deeply, not more simply.

There’s also a deeper shift in how we relate to others. Friendship without an embodied other can lead to submission to a disembodied power. And that power is not spiritual—it belongs instead to systems, images, and invisible structures that shape perception.

This creates a kind of illusion: a world of images and representations that feels real but lacks depth. Within that illusion, a subtle form of narcissism emerges—one that feeds on self-curation and constant mediation.

🔍 Resources for Learning CS

→ DL, DataViz (MIT OCW) & MIT Learn Resources

Learn the core ideas behind neural networks, and explore a broader path of free courses and resources on MIT Learn to build your skills and launch a career in AI and machine learning. If you’re more interested in data visualization, this other course will teach you how to design and code clear visualizations that help people make sense of complex data culminating in a project focused on real-world social issues.

→ Architecture Decision Record

Maintaining ADRs is a great practice, and it has become even more valuable with AI. Martin Fowler’s piece is an excellent primer, and it also links to useful tools.

🔍 Resources for Teaching CS

→ Helpful resource for AP Cybersecurity

An introductory college-level course in cybersecurity that covers threats, risk, and defense across networks, devices, and data, while building skills in analyzing, mitigating, and detecting attacks. Read the full document here.

→ Better Feedback

An episode on how to give and receive feedback. Here’s the audio version and the transcript.

→ SIGCSE 2026 Recap: MongoDB Session Presentations

Kim Yohannan from MongoDB shared two presentations with me on NoSQL design and building RAG/agent apps.

→ Iterative AI Coding Workshop

1.5h workshop idea: demo Claude Code chat on a Tone Matrix task, then have students iteratively build/debug a similar project emphasizing multi-step prompting, testing, and code review. Materials: Assignment, html, GitHub.

→ CSE 373 Spring 2026 - University of Washington

Kevin Lin is rethinking Advanced Data Structures through design-focused learning and agentic coding projects, highlighting how AI is shifting value away from traditional skills and toward architecture. In some of his materials, he mentions UC Berkeley professor Josh Hug, whom I wasn’t previously familiar with but who offers an updated YouTube channel and highly valuable materials for anyone teaching data structures.

🦄 Quick bytes from the industry

The spirit of Foundation is back. New episodes on the live interview show hosted by the great Kevin Rose. Trends, up-and-coming founders, and the technologies worth watching before they become “the next big thing.”

If you’re curious about Uber’s earliest technical challenges, this episode featuring their first CTO, Thuan Pham, is a must-listen. It dives into how the team avoided major breakdowns (like the Dispatch rewrite), expanded into China, rebuilt the app with Helix, and why they ended up creating so many internal tools when open-source options couldn’t keep up.

🌎 Computing Education Community

Dr. Liat Nakar (with Havana Rika & Moshe Leiba) invites submissions to a Frontiers in Computer Science special issue on cognitive processes in AI-mediated problem solving (abstracts May 17; papers Oct 23).
EduCHI 2026 (May 20–22, Toronto, hybrid) is now open for registration.
Paid Summer 2026 GenAI program at NC State (EXLAIM): build real AI tools for K–12 education, gain hands-on experience, and get mentored. Apply by April 7.
The AIED 2026 Workshop on Intelligent Textbooks (Seoul, hybrid) invites research on LLM-enhanced, adaptive, and interactive digital textbooks (shared by Isaac Alpizar-Chacon).
Upcoming Raspberry Pi webinars: Kathryn Jessen Eller on evaluating AI in healthcare (Apr 14) and Shuchi Grover on K–12 data & computing competencies (May 12).
SIGCSE TS 2027 panel “It Seemed Like a Good Idea at the Time” seeks honest stories of failed teaching experiments (shared by Dan Garcia).
UKI SIGCSE journal club (May 11) coming up and University of Manchester is hiring a teaching-focused CS lecturer (shared by Duncan Hull).
SIGCSE seeks self-nominations for a junior coordinator for the 2027 Department Chairs Roundtable. Apply by April 17 by submitting your materials to chair@sigcse.org.
A two-part webinar (Apr 16 & 23) will cover guidelines for responsible AI use in STEM education research, focusing on ethics, research integrity, and protecting participants. Thanks for sharing, Monica McGill.
Illinois CS Summer Teaching Workshop (June 10–11, virtual, free) invites 300–500 word abstracts on CS education (incl. AI & pedagogy)—submit by April 15.
Study on student-centered syllabi seeks faculty input (shared by Nadia Najjar and Debarati Basu).
Software engineering students: take this 8–12 min anonymous survey on learning preferences and AI tool use to help improve SE education (shared by Jeffrey Carver).
Has anyone set up a Rust kernel within Anaconda to keep everything under one umbrella for teaching? If so, Tony Ruocco would love to hear about your experience—feel free to email him at aruocco@rwu.edu.
RESPECT 2026 Doctoral Consortium (June 8, in-person) invites PhD students in computing education & broadening participation—apply by April 8 (shared by Earl W. Huff).

🤔 Thought(s) For You to Ponder…

Related to last week’s topic, this episode of Pausa on imperfection featuring the scholar Diego Garrocho is highly recommended (he’s always a pleasure to listen to).

He already uses oral exams that build useful skills and also help deter AI use (though I don’t think this works for every field).
He also explores the value of difference in contrast to the standardizing homogenization of citizens, why we’re increasingly becoming more alike (in WhatsApp conversations, in teaching, offering a single, uniform model of virtue, etc.).
He also touches on how we’re starting to fake mistakes to seem human; how grace escapes standardized control; and the importance of regularity (knowing there are things that will always be there, and that this consistency holds).
To measure quality, we need an ideal to compare against.
He also reflects on why the laughter in the song “Roxanne” feels like pure truth, or the rebound effect young people experience when exposed to noise without technology.
We need role models: he invites us to critically ask who we’re following.
Imperfect lives generate morally valuable experiences, like forgiveness.
At the same time, this isn’t about romanticizing error, but about moving with regulative ideals aspiring to things that are genuinely good, which also serves an important function.
It’s very difficult to navigate an imperfect world if you don’t have a target to aim at.
AI has no scars, nor does it feel pain from what happens to it.
We talk a lot about humanism, but less about what it actually means to be human (its essence).

This El Hilo podcast episode reflects on the use of AI as a weapon in light of Anthropic’s refusal to work with the Pentagon. Beyond the Anthropic case, I’m interested in what Marta Peirano says about the crossroads these AI companies face as they gain more and more power: either I shut you out, or I fully absorb you. Marta wraps up by saying: “I’m optimistic because I know our problem isn’t a technical one. What we need is a generation of political leaders who are up to the challenge.”

I did theater back in my college days. I really admire stage actors for their public service, their ability to do the same thing every day in a different way, and to create real emotion (to make people laugh, for example). It’s great therapy.

I agree with Josema:

AI has nothing on theater. Theater is something alive that will always endure. That’s why I believe theater is pure magic.
Why should people go to the theater?
Because it’s therapy for getting through life. It makes you a little happier. It makes you feel good. It brings back your optimism, your desire to love and to get excited about things.
Laughter has tons of benefits—for your cardiovascular system, for your brain, for loving more… for everything. It’s something wonderful. When I go as an audience member to a movie or a play and I laugh—and I do laugh, of course, many times—I walk out healthier, happier, more optimistic.
I think laughter is amazing, even if only for the sake of laughing.
What makes a play successful?
Hard work, consistency, authenticity, confidence, and believing in what you do; passion, respect for the audience (who are your coworkers), and talent (craft, stage experience). But luck starts at seven in the morning.

Concha Monje is a robotics researcher at UC3M. I really liked her academic perspective in this Pausa episode on how realistic and viable it is for humanoid robots to become part of our daily lives in the near future—and why. It’s a great listen for setting expectations:

She used to be skeptical, but AI has made her rethink things. She’s seeing major progress in locomotion and motor skills. That raises an interesting question: do we actually want to delegate things like making the bed, cooking, or even escaping our daily routines to a robot? It’s still unclear whether having humanoids in the home would truly satisfy us—or what new needs that might create.
A robot’s ability to move in a stable, human-like way comes down to both hardware (motors, actuators, etc.) and the “intelligence” that tells those actuators how to move. Kinematics have become much simpler thanks to AI-driven automation. That said, getting robots to function like humans in any environment is still a major challenge. They need extensive training and must be able to handle uncertainty in unfamiliar settings—which is why they tend to fail when placed in homes they don’t already know.
Right now, with companies like 1X, you’re not just paying for the robot (whether via subscription or purchase), but also for a human operator controlling it in your home. It’s invasive and expensive, but there’s still a learning process that depends on humans. Unlike industrial settings—where everything is predictable—every home is different, and robots have to figure out how to adapt.
As for strengths and weaknesses: they’ve basically nailed kinematic and dynamic control—walking at different speeds, maintaining balance, coordinating joint movement, all very human-like. What they still struggle with are more complex tasks that challenge balance, like walking while grabbing an object, or moving with the kind of intuitive judgment humans have (for example, picking up a cup—not just recognizing its shape, but understanding how to handle it naturally). That kind of integration into real household dynamics is still missing.
She also shares a great example of robotics being used to help people with ALS, which is genuinely impactful. There are also use cases aimed at addressing demographic and social challenges. That said, she draws the line at replacing therapists—something that opens up a whole ethical debate. There are also security concerns: vulnerabilities through networks, data, or even teleoperators being introduced into the home.
In the end, Concha admits she wouldn’t bring a robot into her own home—she’s had more than enough of them in the lab.

In this week’s column, JL Antúnez writes:

A company doesn’t die when it runs out of cash—it dies earlier, when it loses the shared story of what it stood for.

What are products for, beyond their function? They shape what we see as normal, act as tools of prosperity or poverty, and ultimately transform us as a species.

I really loved learning about the role of a hospital orderly on the A Vivir podcast. I barely knew anything about it before, and it’s such an essential job. It takes a lot of empathy, and they develop a strong intuition—a real clinical instinct.

I’m not a cloud privacy paranoid. I’m perfectly happy paying for Google One and iCloud, among other cloud services but moves like this from Microsoft starting to train its AI models on code from private repositories do raise an interesting debate about how dependent we’ve become on cloud services, and whether it’s worth switching to self-hosting as an alternative, meaning running everything on your own server. It’s not something I’m considering personally, but I wouldn’t be surprised if plenty of people are now crunching the numbers to see what it really costs and whether it’s worth self-hosting instead of paying for these services indefinitely.

Interesting read: The Strength of Weak Ties.

Peter Steinberger acknowledged in his interview with Lex Fridman that it’s natural to mourn the loss of the traditional way of programming (the deep flow of writing code by hand) or the craft but said you can find a new kind of flow by working with agents and building at a higher level. As for the future role of the programmer, he said it’s still about being a builder: the person who defines the vision, the experience, and the key design decisions. In that regard, learning to empathize with agents and direct them learning the language of the agent is the new core skill.

In his interview with Lex Fridman, Peter Steinberger acknowledged that it’s natural to mourn the loss of the traditional way of programming: the deep flow of writing code by hand, or the craft itself. However, he noted that you can find a new kind of flow by working with agents and building at a higher level. As for the future role of the programmer, he emphasized that it’s still about being a builder: someone who defines the vision, the experience, and the key design decisions. In that regard, learning to empathize with agents and direct them, essentially learning their “language”, is the new core skill.

Leo XIV continues his catechesis on Lumen Gentium. This week, he focused on the laity. https://www.vatican.va/content/leo-xiv/en/audiences/2026/documents/20260401-udienza-generale.html

I liked what Aman said about patterns! I see it all the time in my software design students.

📌 Research Corner

We’re making good progress in the ITiCSE Working Group. I’m on the SLR team. Last Friday, we discussed evaluation criteria for papers, and now we are reading them. I’m learning a lot about teamwork and related topics.

This detail in the Claude Code quote is simple, but I had completely missed it. It’s such a great touch.

In our Research Methods class this week, we had a special session on writing. Our guest lecturer recommended Purdue’s Writing Center, and I’d also like to share another excellent resource that has helped me a lot: UH’s Writing Center.

🪁 Leisure Line

Breaking in a new court to play pickleball. I’m getting comfortable with the movement... Now I just need to learn the rules (definitely on my to-do list before the semester ends).

It’s almost a tradition at this point: lunch with Mahdi and a chat about Iran at Cougar Woods Dining Hall. And then enjoying a nice walk around UH campus while praying the rosary.

We spent last Saturday afternoon helping people in need in downtown Houston. Also sharing a rooftop view from my home while getting ready for Palm Sunday.

I’m excited to be reading this new book by HBS professor Arthur Brooks. What I’ve heard so far has made me eager to dig in.

Hot damn, Chick-fil-A sauce is so freaking good. I don’t care how bad it is for me. I think Mahdi likes it too. The one on campus is his favorite spot.

📖📺🍿 Currently Reading, Watching, Listening

Thylacine recorded 200 unique sounds throughout Switzerland to create music. What an art!

Willie Colón passed away this past February. My favorite album of his is El Juicio. It’s the perfect record to introduce to someone who doesn’t know anything about salsa. With tracks like “Timbalero” and “Aguanile,” dancing to the whole album is a serious workout. A huge contribution from Willie to Latin music.

I loved this video. Rocky Trail’s been on repeat for me all week.

A podcast about songs that I’ve been really into lately:

I was recently introduced to Bernardo Bacalhau, and I really like what he shares. An illness forced Bernardo to completely shift his perspective on life. Now he seems really happy in his little cabin out in the middle of nowhere (Norway), after going through the emotional aftermath of his bike journey.

Did a deep dive into Bruno Mars’ new album with this Switched On Pop episode. Perfect time to revisit this beautiful record and just soak it in.

🌐 Cool things from around the internet

A collection of links to stuff I think are worth sharing.

🔗 The work of Derek Lin — these little worlds are captivating.

🔗 IBM Plex — an excellent typeface.

🔗 Sleep Well Creatives — an interactive web-based experience to help people sleep better.

🔗 MegaParse — parse PDFs, Docx, PPTx in a format that is ideal for LLMs.

🔗 Who Is In Space — a website that tells you how many people are in space, what country they’re from, and how many days they’ve been up there.

Issue #38 of Computing Education Things was written while listening to:

🔗 Quick Links

🎧 Listen to Computing Education Things Podcast

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

Buy Me A Coffee

#37 — The Human Touch Is in Style

Daniel Prol — Fri, 27 Mar 2026 14:16:03 GMT

Reflections

Cicero’s view of virtue and eudaimonia. Source: Living Ideas Journal.

I believe in a way of building digital products and doing CS research that respects how humans actually work.

A document released this March by the Vatican’s International Theological Commission, Quo Vadis, Humanitas?, puts a name to something our field has been circling for a while: being human isn’t something we construct or acquire. It comes before us (it’s given!) and it also becomes a task.

Shaping our identity is a lifelong journey, one that unfolds not in isolation but through relationships, time, and genuine encounters with others. This last part sounds abstract until you think about what actually makes a great learning experience. Data structures and algorithms are taught similarly at any school; the topics remain the same. What truly changes is the people in the room learning together. The osmosis of being surrounded by motivated peers, the friction and spark of a genuine intellectual community is fundamental to learning. It is in these spaces that identity is shaped.

That framing matters for Computing Education. Because much of how we design learning systems (and increasingly how we talk about AI!) quietly assumes that the human in the loop is a variable to optimize, a source of friction to reduce, or worse, something to replace.

AI rekindles a sense of wonder about what we can do. But it doesn’t erase our fragility, our dignity, or what it means to be human.

On oversimplification

One of the document’s sharpest insights is that reducing human dignity to cognitive performance or to whatever technology promises to enhance is exactly the kind of oversimplification we should resist. Drawing on Gaudium et Spes, it frames this as a call to discernment: weighing new possibilities against the enduring demands of being human, and asking honestly what fits and what doesn’t. That question doesn’t have easy answers. But Computing Education needs to be asking it.

The soft transhumanism already in our classrooms

The document spends significant time on transhumanism (the belief that technology should help us overcome human limitations, like aging or death). That might sound like science fiction, but a softer version of that logic is already very much present in our classrooms.

When we say, “AI will handle the boilerplate so students can focus on higher-order thinking,” that’s a reasonable pedagogical claim. But when it quietly becomes, “AI will handle the hard parts so students don’t have to struggle,” we’ve crossed into something else. We’ve started treating struggle, often the very place where identity and competence are formed, as a bug to be fixed.

The document frames this as the difference between overcoming limitations through growth and integration versus bypassing them by suppressing or substituting what is human.

Vocation: the human person as gift and as task

The document also raises the question of vocation. The human person is not fully understood as a self-made project, but as someone called to receive life as a gift, to shape their identity with responsible freedom, and to become a gift for others.

From this perspective, humanity doesn’t need an “evolutionary leap” that surpasses its condition. It needs a relationship that redeems it, makes it livable, and elevates it.

For computing educators, this has concrete implications: designing not just for student performance, but for students’ development as individuals: their agency and their identity as people who can build meaningful things with computing.

An invitation, not a solution

The document ends with an invitation rather than a solution. That feels like the right posture for computing educators too.

We’re not going to resolve the tension between AI’s capabilities and human development by finding the perfect policy or tool. We’re going to have to keep asking the question course by course and student by student.

Computing educators will need to keep digging deeper into the relationship between anthropology and AI, trying to understand more precisely the real dynamics reshaping how we know, decide, and relate to one another.

The issue isn’t just what machines can do, but what we’re willing to delegate to them from our own cognitive processes.

What are we actually trying to cultivate? Not just skills or competencies, but people who can think, build, and act with integrity in a world that will look very different by the time they graduate. Only then will we truly be able to accompany students in these exciting moments. That’s something worth taking seriously.

I’d like to publicly thank Javier García Herrería and Giovanni Tridente from Omnes, because this reflection took shape thanks to them. They helped me find the right words and express these ideas with greater precision.

The Vacuum Will Be Filled by Something

An interesting episode full of thought-provoking ideas, though it can come across as somewhat alarmist. As always, it’s worth reading widely before forming an opinion. The title of this episode of the Evangelization and Culture podcast, hosted by Tod Worner with Jennifer Frey, barely scratches the surface — it goes well beyond classic literature and offers a sharp diagnosis of the true purpose of education: a critique of instrumentalism, the role of contemplation, and the state of the academy. It is not a mere utilitarian transaction, but a vital process that shapes identity and human flourishing.

She points to an interesting cultural phenomenon as her starting point: young people, especially young men, spontaneously sharing their genuine enthusiasm for the classics — Dostoevsky, Homer — on social media, completely outside any academic institution. For Jennifer, this is a witness to something that has been lost and needs to be recovered, and also a sign that the basic human need for meaning expresses itself on its own, apart from the educational systems that have stopped feeding it.

Jennifer wants to encourage young people to go further, to genuinely cultivate their capacity for attention. This capacity has been diminished in all of us by current modes of technology, which are often designed in ways that erode our ability to concentrate. She wants to motivate them to develop that capacity and to experience the fruits of that cultivation. Ultimately, it is about leading a contemplative and reflective life, about actualizing our highest human capacities, which are the rational ones. It’s something wonderful. That is why, from within the academy, we need to have the drive to guide young people in cultivating their minds and their humanity, rather than reducing teaching to a mere professional tool.

She argues that we have imaginative capacities, and they will necessarily be shaped by something. If not by great books, great art, or convivial and serious conversation about ideas, politics, and things that matter, then they will be shaped by the algorithm. They will be shaped by pop culture. There is a vacuum, and something is going to fill it.

She believes that for a long time we have failed to take seriously something that was once obvious to everyone: that education is about forming a person. It is not just learning a set of skills to get a job and make money — nobody thought that until very recently. And once you stop to think about it, it becomes clear just how impoverished that conception of education really is.

She gives a concrete example: the governor of Oklahoma issued an executive order to evaluate higher education based on graduates’ salaries and the type of jobs they obtain. And she reminds us that it was Barack Obama who first proposed measuring universities by how much their graduates earn five years after graduation — one of the few bipartisan points of agreement, and precisely one that, according to her, is destroying the classical conception of the liberal arts.

I would also add that at one point the host says “don’t spend too much time on the internet,” and Jennifer compares reading Dostoevsky to using Instagram. I find that framing — as an either/or choice — unconvincing. I think that if we are better educated and more grounded in our values, we will naturally use these tools more wisely. We cannot be naive or ignore how powerful these tools are.

At the same time, Jennifer acknowledges that expertise is valuable and that resources are available. But she does not want to oversell her own role, because she considers it a flawed model of education to think of it the way the Sophists did. She develops this further: for Plato, the central battle across his dialogues is between Socrates and the Sophists. The problem with the Sophists is that they give beautiful speeches, but when subjected to dialectical conversation — where objections are taken seriously — everything falls apart. The true educational model is Socratic dialectic: a conversation where the opposing position is genuinely engaged, and if it cannot be refuted, it must continue to be taken seriously. It is something deeply personal, because in these exchanges the interlocutor’s own flaws come to light.

She notes that reading is necessary, but not sufficient. Discussion and dialogue in community are also needed. That is where truly meaningful friendships can form, if approached in the right spirit. Others can help not only to interpret a text, but also to understand oneself.

She explains that the most formative period in a person’s life is between the ages of 18 and 22. That is when adult identity consolidates, and for many people it coincides with the college years. So the idea of giving up on liberal learning at that very moment and simply saying “well, he’s going to be an engineer” is a disaster. It makes no sense. It does not conform to what we know about human nature. We need to recognize that this is the most critical formative period. It is also when young people are no longer under constant supervision, and that is a good thing. That separation is necessary for forming an adult identity. But you cannot simply abandon guidance, or reduce higher education to something purely external or instrumental.

She connects this to the question of rereading, responding sharply to parents who tell her “Jimmy already read Homer in sixth grade, he’s done with that.” For her, this reveals the depth of the problem: even someone who makes sacrifices to send their children to a classical school does not truly understand what formation means. Reading Homer at twelve is wonderful, but it is not the same as reading it at eighteen surrounded by people with very different worldviews, within a semester that moves from Homer to Cicero, tracing the red thread running through an entire tradition.

She argues that one is never done cultivating the mind. It never ends. And insofar as you stop doing it, you are denying yourself — denying yourself the possibility of human flourishing.

She also observes that nobody talks to students this way about their education. Nobody says to them: “Sure, you could skip the reading, but you’d be denying yourself human flourishing.” Nobody frames it that way. And she says that when you do say it to them, their eyebrows shoot up.

She believes that a large part of the problem lies in how education is presented to young people. Frankly, it is not surprising that they reject it. If you tell a young person they should read Dostoevsky because it is a form of mental gymnastics that might make them a better coder, they are going to think: “there are plenty of other ways to achieve that” — without having to keep track of ten different names for every character.

It is a bad sell. It shortchanges everyone involved. And in her experience, higher education administrators and leaders seem determined to double down on this failure. It is very hard to break through the blinders they seem to have on.

She finds the situation sad. She just finished her annual review and receives no credit for her podcast, no credit for engaging with a general audience, no credit for writing for a broad readership. That work simply is not valued.

She believes this needs to change. She does not think it will, but it should, because it has many negative consequences for everyone.

She concludes that much more attention needs to be paid to undergraduate teaching. In fact, it should be a fundamental priority — but unfortunately it is not in many cases, as there is always faculty focused primarily on research.

On the canon, she responds to those who accuse it of being exclusionary or of being “just dead white men”: she finds this a partially valid but misdirected objection. She notes that a living tradition always evolves and the canon is not set in stone. But the fundamental question is not “is my canon diverse enough?” — it is “are these texts truly great, in the sense that they are timeless and have been foundational to this tradition?” That is what really matters. She also points to something telling: the English major has collapsed, and it is no coincidence — when the literature department stops loving literature and becomes obsessed with the politics of the text, it loses the very students who came looking for precisely that: the story, the narrative.

On AI, Jennifer Frey argues that the experience of a poetic work cannot be reduced to a summary. The truth of the poem is embedded in its literary form and in how it acts upon us, so a summary or an AI cannot offer the full experience of the poem. She adds that obtaining an AI summary and thinking you have experienced the work is simply not enough — poetry shapes the imagination in a way that mere paraphrase or informational extraction cannot.

She also takes on the difficult question Todd raises: what about the cultured Nazi, the monster who read Goethe in the morning and then went to Auschwitz? Her answer is nuanced. First: virtue cannot be “taught” mechanically, because there is always human freedom involved — the student receives, interprets, and decides. Second: Nazism politicized culture in a very specific sense, viewing the greatness of German culture as proof of Aryan racial superiority. That is not reading well; it is ideological misreading. And third: the fact that art can be misinterpreted or abused does not invalidate that it is ordered toward going well. As she puts it, the possibility of failure does not negate the reality that something is meant to succeed.

Finally, the guest rightly notes that many will say humanity has simply moved on and that everything must now be transmitted visually. But sitting down, picking up a book, and spending time reading takes effort — and yet it has at least two advantages (and these I add myself):

When we read, we can follow the thread of an argument and a line of reasoning, and we also train our minds to reason — that is, to move from one thought to another, not merely from one sensation to another. The world of the image is dizzying, and what dominates in images are sensations: wonder, pleasure, anxiety, sadness, joy. The written word forces us into a different dynamic: it forces us to reflect. Consequently, if reading disappears, so does our capacity to construct or discern arguments. It is no surprise that many young people are losing this ability.

Reading strengthens our imagination, unlike the passive reception of images in a video, which come already fully formed.

What other reasons can you think of for why recovering reading matters? Images certainly have their place, but I believe we need to recover reading. If you’re interested, here is what Jennifer recommends reading — with humility, in community, and returning to them more than once to go deeper: Madame Bovary, Dostoevsky (e.g. The Brothers Karamazov), Homer, Virgil, Paradise Lost (John Milton), Plato’s Dialogues (Apology, Protagoras, Gorgias, Phaedrus, Republic), Saint Augustine (Confessions), Saint Thomas Aquinas.

The New Cool Thing: Being Human

Reading? To live, to discover new worlds.

How do you read more? Ask for guidance, spend time in bookstores to spark that desire.

Writing? To generate ideas, to shape who we are, to organize our thoughts.

How do you write more? Practice so you can express yourself clearly.

Listening and engaging in conversation? To stay informed, to develop a clear point of view.

Because having a point of view isn’t the same as being polarized: it’s about understanding and building bridges.

At the end of the day, it’s about cultivating yourself: growing intellectually and culturally. Living with that kind of mindset is becoming more important than ever.

Ted Gioia touches on this in his newsletter, using the example of a bookstore in Alabama that only sells books signed in advance by their authors. I think we’ll keep seeing more initiatives like this, where the human touch becomes a key differentiator in response to the rise of AI tools and the content they produce. As I mentioned in issue 35, human authorship is becoming increasingly valuable.

The Honest Broker

The New Cool Thing: Being Human

A bookstore in Alabama keeps getting covered in the national news. Even The New Yorker recently sent a reporter to visit The Alabama Booksmith, a small, almost windowless business on a dead-end street in Birmingham…

3 months ago · 751 likes · 101 comments · Ted Gioia

🔍 Resources for Learning CS

→ Coding agents for data analysis

This post from Simon Willison goes through everything from a hands-on workshop on agent-assisted data workflows, including querying, cleaning, scraping, and visualization with Claude Code and Codex. It focuses on end-to-end workflows using Python and SQLite.

→ Master Machine Learning with scikit-learn

This new book from Kevin Markham is a practical guide to the ML workflow covering data prep, pipelines, model evaluation, and the habits that drive reliable results.

→ Claude Skill for Deliberate Skill Development

AI coding tools speed you up but can weaken your understanding. This Claude Code skill inserts quick learning exercises into your workflow, prompting you to predict, explain, or build key pieces yourself so you still gain real expertise.

🔍 Resources for Teaching CS

→ Better Meetings

We academics have many meetings, and, as with many things, we were never trained in how to have good ones. This is a great episode for learning how to run effective meetings within the academic system and 1:1 meetings.

Your Meetings Suck. Here’s How to Fix Them.

Unlocking the Potential of 1:1 Meetings

→ GitHub Education Teacher Toolbox

The Teacher Toolbox is a resource hub for educators and administrators to get started with GitHub. Here are more resources for teaching with GitHub.

→ Data Visualization - A Practical Introduction (2nd Ed)

Kieran Healy’s Data Visualization is getting a major update: modern R tools, refreshed examples, and new lessons shaped by years of teaching and real-world use.

🦄 Quick bytes from the industry

→ Vishal Misra (Columbia CS Professor and Vice Dean of Computing and AI) on The a16z Show

Great Andreessen Horowitz AI ep featuring Vishal Misra and Martin Casado on LLMs. They cover Kolmogorov complexity, Shannon entropy, relativity, causal inference, Don Knuth, consciousness and, of course, cricket (Vishal is actually the Dean of Cricket Analytics for the San Francisco Unicorns).

→ Without juniors, there is no future

Musings Of A Wandering Mind

The Pipeline Problem

The industry has a new favourite story. Senior talent plus AI equals superpower. More output, fewer people, better margins. S4 Capital’s Monks just put a subscription model on top of it - ditch the timesheets, pay a flat fee, get the best minds amplified by a machine that keeps getting smarter. Clean, logical, easy to sell…

3 months ago · 119 likes · 18 comments · Zoe Scaman

The tech industry has spent the past few decades leaning heavily on junior talent. When revenues started to decline, big companies moved quickly to cut costs, slashing salaries and laying off a significant portion of senior staff, replacing them with younger and lower-cost workers. Now those same juniors are at risk of being replaced for the very same reason the seniors were: cost-cutting.

That decision is already creating a new set of problems, even if they look a bit different this time around. Zoe Scaman touches on this in her newsletter, specifically in the context of the marketing industry but the insight applies just as much to tech:

In ten years, is not a shortage of operators. There will be plenty of people who can use the tools, hit the brief, generate clean work fast. What will be missing is the people who can tell you if the brief is wrong. Who can sit in a room where everyone is nodding and identify the thing that isn’t being said. Who feel that a campaign is going nowhere before a single piece of evidence exists. That quality doesn’t come from tools or training courses. It comes from time and friction and proximity and failure, repeated until it becomes instinct. And we are in the process of making sure the next generation never gets any of it.

→ The Path to Apple PM (and Why Consulting Wasn’t It)

Aman Manazir sat down with Nina Wang, an Apple Product Manager, who went from hating her consulting job at Strategy& to landing a PM role at Apple after one of the most brutal job hunts she’s ever been through.

They talked about:

Why she chose consulting over PM straight out of MIT and regretted it within a month.
The real difference between software engineering and product management.
Exactly how a software engineer can pivot to PM.

They make a really good point about combining multiple skills to stand out. That’s something I personally relate to: content creation, podcasts, newsletters, social media, and so on. Figure out that unique mix that sets you apart. This is that part of the conversation:

Nina:

I’m someone who I like to be the best at something. If I’m not the best, then I try to win at a different game, if that makes sense. Like, if I can’t win the game we’re playing, I’ll find another game I can win.

Aman:

I actually feel like I’m very similar. Pretty soon, I realized I had kind of maxed out—not in absolute terms, but in ranking. I was never going to beat the math genius in math. So I thought, “Ok, it’s time to start a YouTube channel,” because that’s a completely separate game.

That’s a game where the math genius, who might not have strong communication skills, can’t compete in the same way. So it became about combining all these unique aspects of myself. Even within YouTube, there are already a lot of STEM, CS, and software career creators, so I asked myself: what other interesting parts of me can I bring in? I’m really into self-improvement, productivity, health, and lifespan extension so how can I incorporate those into what I’m building? Did you ever feel like that—like you were crafting your own “category of one”?

Nina:

I think that’s actually what everyone should strive to do. Most people are never going to be the best at any one thing and that’s not really what matters. What matters is all the other facets of yourself that you can bring to the table. That’s what people truly value.

I think that applies to almost any scenario whether it’s a job, friendships, or getting into a school. It’s all holistic. You might not have the best technical skills, or the best presentation skills, or be the funniest person but you have your own mix of interests, strengths, and even weaknesses. And together, that’s a package only you can offer. As cliché as it sounds, there really is only one of you so you might as well embrace that.

Aman:

I feel like there are only a handful of people who are truly the best at something, and hats off to them they’re number one. But they’ve usually dedicated their entire lives to that one thing. I mean, we’re in the middle of the Winter Olympics right now those people have been training since they were kids. And even among all the people who’ve trained their whole lives, only one comes out on top.

Speaking of combining multiple areas, I have a friend who’s a software engineer at Microsoft. She’s also built a big platform on LinkedIn and Instagram, and they actually chose her to be a spokesperson for Microsoft Copilot. So she’s both a software engineer and a public-facing figure, and she’s even contributing to a book with the CTO of LinkedIn.

I was talking to her about it, and she said she knew she would never be the number one software engineer. She thought it was cool, but she wasn’t going to beat everyone else. So instead, she leaned into her strengths in media, fashion, and presentation and that’s how she carved her own path.

The more I see examples like that, the more I feel like people who lean into their unique combinations and build their own “category of one” tend to be the most successful.

Nina:

Yeah—and not only more successful, but it’s also just cooler. Life is better that way. I mean, being the best software engineer of all time is impressive—but it’s mostly just that: impressive. But if you’re a software engineer and you’re into fashion, or music, or something else—and you build something around that combination—that’s way more interesting.

I also liked Nina’s take on why to pursue a hybrid major (multiple fields/disciplines), beyond just the fact that it sounds cool, fits an indecisive personality, opens up more options in the job market (you diversify your portfolio and don’t depend on a single career path), or reflects not being able to compete in a pure CS track. It’s about looking at problems through different lenses.

That kind of generalist profile tends to work better with people in business, and you can benefit from taking classes at a business school—getting exposure to great business professors whose classes are often more engaging. It gives you a broader experience, though with less depth in the technical foundations of software engineering. It’s a tradeoff.

But as Nina says, if you’re passionate about one thing, go do that.

Software engineers who want to become PMs are those with technical fluency who can translate engineering constraints into product requirements, strong communication and presentation skills for cross-functional collaboration, good decision-making that balances trade-offs, and domain knowledge or specialization to provide context; the easiest path is to pivot internally where you already have institutional knowledge, and you should practice “product thinking” by reframing your engineering work around user problems and business goals while highlighting past product-impacting projects when applying elsewhere.

Advice to Freshman-Self: Leverage the resources and people around you—make connections, dig deeper to find common ground, and learn from those networks—while remembering that achievements and prestige aren’t everything, so don’t build your life solely chasing status; prioritize meaningful relationships and personal fulfillment instead.

→ Inside Kubernetes: Brendan Burns on Building, Scaling, and Surviving Big Tech

Brendan Burns is the co-creator of Kubernetes and current technical fellow/CVP at Microsoft working on Azure. Recently Ryan Peterman interviewed him for all the stories behind building it. They discussed:

How they convinced Google to open source it.
Why he says you should hide ~10% of your bandwidth from management.
The hardest technical challenges.
Scaling up Kubernetes for AI workloads.
Top technical book recommendations.

Is a PhD worth it? Brendan’s perspective:

I had a lot of fun. I really enjoyed doing a PhD in robotics, so that alone made it worth it to me. Secondly, I think I learned a lot, especially from my PhD and my advisor about how to write and present my ideas, both in written form and in presentations. Those are skills you don’t necessarily learn in industry, and they’ve benefited me a lot. I also improved my ability to argue effectively. For example, we talked about that six-month period when we were arguing for why we should be allowed to open source something. The skills I developed in writing and presenting really helped during that time and have continued to help me since.

Later, when I worked as a professor for a couple of years teaching CS101, I had to explain concepts to students who didn’t know much about computers. That experience helped me organize the early parts of the Kubernetes project in a way that made it easier for others to learn. People would come in asking, “What is a container? What is orchestration? How do I do this?” There was a lot of teaching involved. My experience as a professor—thinking about how to teach effectively—really helped me do a good job of introducing Kubernetes to people. I learned a lot of useful skills that significantly benefited my career—and I had fun along the way.

In terms of what you should learn, Brendan says:

I actually don’t really care what you learn—I care that you’re learning. The most important thing is to find something you’re excited and energized about. That’s what you’ll choose to do instead of, say, watching YouTube.

If you’re not excited about AI, you’re probably not going to do a very good job learning it, which means you’ll end up wasting your time. But if you’re really excited about systems, you’ll likely put a lot of passion and energy into it—and we still need systems engineers.

I think this is a pretty common question. I sense that there’s a lot of fear around making the wrong decision. I always tell people: I never had a plan for my career—never, ever. I’ve just chased things that I found useful, fun, and interesting.

Of course, that approach can work out badly for some people. Having a plan is probably a good idea for many. But I also want people to understand that, in hindsight, things that seemed like mistakes or dead ends often turn out to be critical learning experiences.

So worrying about whether you chose—or will choose—the wrong thing isn’t that helpful. As long as you’re learning, you’re probably doing ok.

Books he recommends

Design Patterns: Elements of Reusable Object-Oriented Software

Leadership on the Line

The Five Dysfunctions of a Team

Brendan says he would tell his younger self to keep better notes because there’s a rich story and many interpersonal details from the Kubernetes journey that he now wishes he’d recorded for a future write-up or book.

🌎 Computing Education Community

UC Berkeley’s CS KickStart student organizers launched a Slack to connect institutions running summer bridge programs and share resources.
ACM TechTalk (Apr 9): Angela Sara Cacciapuoti explores the architecture and protocols of a quantum-native Internet.
NSF CyberAI SFS seeks reviewers for 2026 panels (May–Dec); complete the survey to join and contribute to AI & cybersecurity education funding decisions.
ICER 2026 Doctoral Consortium (Uppsala, Aug 11) invites PhD students in computing education. Thanks, Briana Morrison, for sharing!
PhD study seeks CS instructors for a 45–60 min interview on student transfer between programming languages. Contact Gul e Fatima Kiani (gkiani@unomaha.edu) or Dr. Brian Dorn (bdorn@unomaha.edu).
Opetunde Ibitoye (University of Cincinnati) seeks students/recent grads for a 10-min anonymous survey on how they use LLMs for learning.
Bates College seeks a Visiting Assistant Professor in Digital & Computational Studies (start July 2026)—teaching-focused, interdisciplinary role with R/Python courses; apply by March 31.

🤔 Thought(s) For You to Ponder…

Marlene Mhangami (Microsoft) shared a practical ACM TechTalk on agentic coding workflows. GitHub repo.

Joaquín Cuenca (Freepik) in El Diario on democratizing creativity. Freepik is based in Málaga and ranks #11 among the most-used AI websites in the world (via a16z).

Our thesis is that as AI evolves, it’s going to build an entire industry around creativity. Imagine you’re a creative and you want to make a movie. Until now, that was almost impossible. You had to convince a lot—a lot—of people to get an indie film made. And it required a major investment—we’re talking tens of thousands of euros. People couldn’t really have filmmaking as a hobby. They couldn’t afford to tell stories in that medium. But now, they will be able to. And that completely changes the landscape.

People tend to prefer AI-generated text because they can’t tell the difference but if you label the author, they lean toward the human-written version. As Max points out in his newsletter, these findings don’t validate either side, but they do make one thing clear: it’s getting harder and harder to tell whether a text is real or digitally created. Still, people prefer human-made content when they’re aware of it.

Read Max

What do "which is A.I.?" quizzes tell us?

This newsletter is brought to you by Squarespace…

2 months ago · 119 likes · 26 comments · Max Read

Apple too! I don’t think it’s just another feature. It actually signals a clear trend: the blending of audio and video in podcasting. It raises the question: what does “podcast” even mean today? It’s no longer just something you listen to as part of your daily routine. We’re not only tuning in while cooking, working out, or driving. It’s increasingly becoming more like a talk show adapted for the internet. I’m not a fan of extremes, either rejecting video altogether or clinging to some idea of “pure” radio. I just hope there continues to be high-quality podcasts that preserve that intimate listening experience.

“When the line between what’s true and what’s false disappears, it doesn’t create chaos but something more sophisticated: a longing for that distinction. The value of what’s authentic gets inflated.” José Luis Antúnez in his newsletter.

If I lived in London or anywhere in the UK, I’d definitely attend the Christian Art Conference.

It’s pretty cool what you can build with the Claude Agent SDK (a favorite tool among a lot of devs for creating agents). This use case is especially interesting: building a pool coach.

I was really struck by the images from the Vatican’s new Stations of the Cross, and I ended up spending quite a while mesmerized on the artist Manuel A. Dürr’s website.

I found this Spanish “footnote” by Teo Peñarroja on Simone Weil, the future, and desire really interesting.

What are the world’s deadliest animals, and can we protect ourselves against them?

Gema Pérez is a scholar at Valladolid University, but she’s also a film critic. This episode about the movie Barbie is excellent, and I highly recommend it if you’re interested in the themes that come out of the film.

That said, what I really want to focus on are two other reflections:

The first has to do with the role of academics in relation to cultural impact.

For me, it’s genuinely exciting that it reached such a wide audience because, at its core, it helps make contemporary culture—and the big questions being discussed—more accessible to the general public. Those conversations enrich all of us. We need to find answers and learn how to recognize them so we can engage meaningfully with the world we live in. This really highlights the value of public scholarship: making ideas—and the people who work with them, like university professors—accessible to society, which is ultimately the purpose of what we do.

The second reflection is about taste:

It’s more about loving film than just consuming it. It’s the difference between someone who savors wine and someone who just gets drunk on cheap mixed drinks at the beach. There’s a big gap between people who know how to appreciate what they’re watching and those who don’t. We live in a society overflowing with audiovisual content, and not all of it is created equal. The key is to watch high-quality work, get familiar with the classics, and develop visual literacy. That’s what enriches you, because over time, it allows you to tell the difference between a fine vintage and a cheap buzz—and to truly appreciate great storytelling.

It’s about sharpening your eye, developing your sensitivity, and letting yourself be guided by people who know what they’re talking about. What happens with film applies just as much to literature, music, or painting—it’s the same process. But you have to want it. Sometimes we associate television with mindless consumption, and that completely dulls the spirit. It cuts us off from beauty, from greatness, and from the power of cinema as an art form.

Fulton J. Sheen will be beatified in St. Louis on September 24. He was a media pioneer and left behind a profound spiritual legacy in the United States. He was a communication genius, much like Barron. Both Sheen in his time and Barron in his have been tremendous forces for American Catholic pride.

ACM Fellow Fred Chong highlights how advances in quantum hardware, error correction, and software are bringing us closer to practical quantum computing as a complement to classical systems:

Once these physical qubits can implement 100s of reliable logical qubits, quantum advantage on a range of applications will be enabled.

HPC is critical to solving real applications with quantum computers.

📌 Research Corner

Yesterday, I deployed the new tool I’ve been building for my PhD research project to production for the first time. So far, things have been going well. I haven’t encountered any login issues. I’ve also been fixing a few bugs on the assignment page as students report them.

I’ve been providing one-on-one feedback on Canvas based on how students are handling the assignments. I can now generate feedback much faster and tailor it to each student, since I have access to their sequences, guesses, and number of attempts in the dashboard/database. I’m also using Claude Code to quickly draft feedback, which I then review and humanize.

Camilo introduces an experiment here showing that when multiple AI agents generate parts of code without communicating, they tend to produce incompatibilities due to a lack of shared specifications and the intrinsic cost of coordination—what he calls the “specification gap.” The problem is both technical and epistemic: systems need to be designed to make explicit what agents must know before they act, because silent failure—not realizing that something is going wrong—is especially dangerous.

🪁 Leisure Line

I absolutely love Spain’s new jersey for the World Cup kicking off in June. It’s super clean and really well thought out down to the smallest details.

Last night just wasn’t our night. Tough ending in March Madness for Houston. I really thought they were going to make the Final Four again this year, but credit to Illinois—they played a great game and earned that win. Congrats to them on a great season. And I’m still really proud of the Cougars…

📖📺🍿 Currently Reading, Watching, Listening

This animated short film by Lorenzo Fresta is already over five years old, but it’s beautiful, with a style that feels like anime from the ’70s and ’80s.

Season 5 of The Chosen is a great opportunity to dive into the upcoming Holy Week.

A great conversation between Fermín Torrano, a war correspondent in Ukraine, and David Jiménez, former director of El Mundo, offering a behind-the-scenes look at war reporting.

What Fermín’s mother told him before he left for the war: “Whatever makes you happy.” The same thing David Beriain’s mother said in an interview after he was killed in Burkina Faso: “That’s what made him happy, and it wasn’t my place to tell him not to do it.” As David Jiménez puts it: “That’s an extreme kind of love and a brutal kind of generosity.”

Why do some movies feel alive, while others don’t? It often comes down to perceptual realism.

💬 Quotable

You CANNOT increase your joy by avoiding discomfort. Joy and pain run on the same neural circuits, so you actually decrease your capacity for joy by avoiding discomfort.

― Megan Cornish

🌐 Cool things from around the internet

A collection of links to stuff I think are worth sharing.

🔗 Ritmos Latam — some of the rhythms you hear across Latin America.

🔗 All SVG icons — search 250k+ Free SVG icons.

🔗 What the War Has Done to Iranians — the New Yorker chronicles the Iran war through WhatsApp messages.

🔗 Ziptable — share small datasets as a link. Made by Evan Peck.

🔗 Data Kids — visualization contest for children.

🔗 Recordly — create polished screen recordings in seconds.

🔗 J-Card Generator — the fastest way to make an insert for your cassette tape cases.

🔗 Impeccable — design fluency for AI harnesses.

🔗 Deepstash — microlearning app designed to replace doomscrolling.

Issue #37 of Computing Education Things was written while listening to:

🔗 Quick Links

🎧 Listen to Computing Education Things Podcast

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

Buy Me A Coffee

#36 — Delegating Understanding

Daniel Prol — Fri, 20 Mar 2026 14:47:08 GMT

Before we dive in, a quick personal note: I’ve been accepted by the Aalto Science Institute at Aalto University as a Visiting Doctoral Researcher for Summer 2026, and they’re funding me. I’m so excited to be in Helsinki in June and early July with Juho and the rest of the group!

Reflections

The Cost of Delegating Understanding

Isiwal, CC BY-SA 4.0 https://creativecommons.org/licenses/by-sa/4.0, via Wikimedia Commons

Let me kick things off today by sharing this piece by Miguel Ávila Monroy on the idea of the delegation of understanding in today’s AI context. Miguel writes about our tendency to hand off the task of thinking for ourselves to a third party. I kept turning this idea over in my head because it describes, with striking accuracy, something I see all the time in computing education. We’ve never had more access to knowledge, and yet we delegate understanding with astonishing ease. At the slightest doubt, we turn to something else to decide for us, not because we can’t figure it out ourselves, but because we’ve come to assume that this third party will always reach a better conclusion than we would.

In computing, this becomes especially visible. A student who never truly debugs a program and who jumps straight to asking an LLM to fix it, learns to follow the map. But the map is not the territory. Knowing that bubble sort compares adjacent elements is very different from having implemented bubble sort, then insertion sort, and feeling in the numbers why one takes seconds and the other milliseconds on a list of ten thousand elements. The same happens with Prim and Dijkstra. You can know that both are greedy algorithms without ever understanding why they are not interchangeable, until you sit with the same graph, run both by hand, and feel the moment where their goals quietly diverge.

That time, which may look unproductive, is where something real actually happens. Truly understanding something changes you. You move from watching where you place your foot to beginning to see the landscape.

And with that often comes something unexpected: an appreciation for the material itself, and a sense of wonder at those who move through it as if they were floating. The question we ask our students (and perhaps the one we should ask ourselves) is not whether they have the ability to dive in. It’s whether they’ll choose to use it.

On Perceptions in Academia vs. Industry

I’ve been asked several times why I decided to switch to academia after working in industry. I connected with this post by Andrew Kyle Lampinen, which raises the broader question of where research is better pursued, academia or industry:

Infinite Faculty

On research careers in academia and industry

I get asked a lot of questions about the relationship between AI & Cognitive Science, especially from early-career researchers wondering where their work might fit into the rapidly evolving fields. This is the final in a series of posts where I aim to lay out my current thoughts on the relationship between these fields — and the career options in them …

4 months ago · 39 likes · 5 comments · Andrew Lampinen

For context, before moving into academia, I spent 10 years in industry, from startups to larger corporations. Now that the academic year is coming to an end, my first year as a PhD student, it feels like a good moment to reflect on which environment better aligns with my philosophy of life, as well as what I appreciate about each.

Unlike the author, I have more experience with startups, but I still found the framework relevant. I adapted some of the categories to my own context and perspective after this past academic year.

Mentorship: In academia, mentorship flows in both directions and is constant. I learn a lot from my advisors and fellow PhD students. As a TA, I also mentor undergraduate students and sometimes other PhD students, although it often feels less like formal mentoring and more like being a supportive peer and sharing what I’ve learned. In industry, mentorship from managers or peers played a smaller role, mostly because of the pace of work. That doesn’t mean the relationships were not strong, but the role of mentorship was different.

Diversity of backgrounds: I learn a lot from my PhD classmates, which is one reason I prefer attending classes in person. Unexpected conversations often come up. I also learn from my research group, which is very diverse, and it is useful to compare perspectives from Europe, Asia, and the U.S.

Advocating for one’s work: “In both academia and industry, part of the job is advocating for your work.” I agree, and I regularly use what I learned in industry. It is transferable.

Work–life balance: This was not an issue in my last company, where I had a flexible schedule. That said, I have generally found more flexibility in academia, even though it often requires more in-person presence, which I am fine with. For example, during Thanksgiving week and Spring Break, I did not work, and that is acceptable as long as your advisor agrees and you meet deadlines. This path allows me to spend more time with my family than the startup world did. I also have friends in large corporations whose lives are more stressful than mine, although there are busy periods here as well.

Stability: In my case, the salary difference has not been large, although I know that in the U.S., the opportunity cost of leaving industry can be high. As long as I can live comfortably, I am okay with that and hope it will pay off later, either in academia or industry. UH has also supported conference travel, and I have volunteered at some events to avoid registration fees. Securing funding is part of academic life. Sometimes it depends on me, and sometimes it does not.

These reflections are not especially deep or surprising, but they reflect my experience so far. I am sure I will have more to add as I continue this academic journey.

The Joy of Reading

Ricardo Calleja reflects on freedom, reading, and the kind of inner life that books make possible in this episode with Helena Farré Vallejo.

The conversation moves from our modern disillusionment with freedom to the idea of reading as a vital priority, not just a pastime, but a way of slowing down, thinking more deeply, and understanding life with greater clarity. Along the way, Ricardo talks about intellectual “diving,” the moral and formative value of reading, the difference between essays and novels, and why the physical book still matters so much. It is a rich conversation about attention, depth, and the search for a fuller human life. Quoting from the podcast:

Freedom

One of the first chapters of the book is about our problem with freedom. I say something that seems quite obvious to me: we idolize freedom while at the same time being afraid to use it and questioning whether we are truly free. That was already true ten years ago, to give a rough estimate.

But now another attitude has appeared: we have become disillusioned with freedom. We have tried everything and found that it does not really satisfy us. I think that is the state we are in now.

Everything I said before still applies, and even if it sounds contradictory, we idolize something we distrust, something that frightens us. But above all, I think our dominant attitude now is disillusionment. And all of this makes us less free. It makes it harder for us to venture into freedom.

Reading as a vital priority

I love reading. In fact, I would love to read even more. It is not that I do not read, but for me, the definition of a better life is a life in which I have more time to read.

When different options and decisions come my way, one of the criteria I use is: will this allow me to read, or will it prevent me from reading altogether?

In that sense, reading is a difficult activity, and I think this is true for many people, not just for me. But yes, I do like reading.

When I am actually able to read, it is a sign that things are going well. It means I am not completely overextended, that I have not gone past my limit. It means I am able to stop the hamster wheel, which is not always possible.

Intellectual diving

Reading, in a very evident way, especially today, allows us to stop and reflect, and to do so more deeply. Thought and reflection have layers of depth, and sometimes we think that once we are on the surface layer, that is enough. “I already think.” Well, yes, you think something, unless you have been totally lobotomized by social media or novelty. But just as with interior silence, there are levels.

As you descend through those levels, you discover a world with greater clarity and more nuance, a world in which everything connects with everything else. Reading is practically the only way we have to make that dive, to go downward.

And like every dive, it means entering an environment that feels somewhat unnatural, difficult, something that makes you want to come back up, see the light, and breathe. But little by little you discover that there is a world down there, and that it is worth staying there and spending hours there.

Reading to understand life

What I look for is understanding. Understanding what? The world, life.

I am drawn to authors who offer a vision of life, not something obvious, not a repetition of commonplaces, but something that shows how the pieces fit together at a level you did not expect.

Reading is, in the end, a search for something more. I think we all have this: dissatisfaction with present life, dissatisfaction with the explanations we have at hand, and the intuition that there is someone who can accompany us better through life, someone who has understood us even before knowing us.

Sometimes books put words to things for which you had no words. Other times, they repeat, almost letter for letter, something you yourself had already sensed. And then you think: how is it possible that this person is saying the same thing I say, not the same thing everyone says on the news every day, but the same thing I say that nobody says?

Reading and moral improvement

The classical tradition spoke of moral virtues and intellectual virtues. Reading demands a certain moral virtue, especially today, because without some degree of self-control, you simply do not read.

But fundamentally, what reading contributes are intellectual habits. Does that make you a better person? Not directly. Precisely the distinction between intellectual and moral habits means that the answer is no. Reading does not make you morally better by itself, because it is not directly a moral habit.

But it does make you grow in an important dimension of being a person, one that is necessary for the full development of the moral virtues, even if it does not automatically bring them about.

Culture, reading, all of that can be misused, obviously. History gives us more than enough examples. And yet reading can also shape us morally. It has always been used that way, educationally and even for self-education.

Good literature cannot be moralistic in the simplistic sense, because in life two things happen. First, not everyone who acts well ends up doing well. That kind of moralism, “the good person prospers and the bad person fails”, is false to life. It is bad literature because it is bad biography and bad morality.

Literature helps us because it allows us to carry out, as it were, an experiment in advance: what would I do? It does not usually ask the question in such a direct way, but as readers we ask it implicitly.

Essay vs. novel

I now have the personal, theoretical, and experiential conviction that I get much more out of a good novel than out of most essays.

The problem is that it is much harder for me to hold my attention on a novel, because my imagination wanders. With an essay, unless it is an extremely complex or fundamental book, I feel in control. I take the book, look at the index, go here and there, skim, stop, underline, skip sections. I do whatever I want with the book and extract what I want from it.

A novel is different. A novel demands a pact of fidelity from the first letter to the last. And you do not know whether you want to make that leap of commitment.

It is not so much fear of commitment as fear of wasting time. I am going to dedicate hours to this, it had better be worth it.

The physical book

I like reading on paper, with a pencil in hand. I do read more on Kindle now, especially because I travel, and I underline there too. But it is very different.

Reading on a Kindle is still reading, of course, but the physical book represents something important: when I read, I am doing a physical activity, and my memory is physical too.

With a physical book, I can feel where I am. I can see that I have reached the halfway point, that I have a little less left, that I have crossed a threshold. That may seem like a trivial thing, but I do not think it is secondary to the reading experience. It is part of it.

And then there is the visual memory of the page. I more or less have a photographic memory, so I can often remember where a passage is depending on its place in the book: the height on the page, the shape of the paragraphs. With Kindle, all of that disappears. It becomes like quicksand. Every page is the same.

Books also matter as objects. Their presence on the shelf matters. Even the books I have not yet read but would like to read are there as a challenge, as a reminder that there are still many worthwhile and delightful things waiting for me.

Some books mentioned in the episode: Mrs. Frisby and the Rats of NIMH; East Wind: West Wind; White Fang; Moby-Dick; The Neverending Story; The Lord of the Rings; Memoirs from Beyond the Grave; Dune; 84, Charing Cross Road; The Road; Hamnet; Laurus; Confessions; 247 Days; The Tartar Steppe; The Ides of March; The Pearl; The Martian Chronicles; The Brothers Karamazov; A Grief Observed; The Marble Cliffs.
If you’re looking for something technical to read, I put together a list of essential tech books in English and Spanish in this Google Doc some time ago. Feel free to check it out.

🔍 Resources for Learning CS

→ Do you want an AI from the Stanford d.school to audit your project?

Scopey asks a few questions about your project to help refine its scope and goals. This chatbot is designed to guide initiatives toward a human-centered design approach.

→ Measuring the quality of AI-generated code

There’s a concern going around among software engineers who use AI: how do I know the code I’m generating is actually good? Kimun is a command-line tool Eduardo Díaz built to objectively measure code quality using metrics backed by decades of research: Halstead metrics, cyclomatic complexity, and Adam Tornhill’s ideas on temporal analysis. The best part: it integrates directly with Claude Code. You use AI to generate code, and then use AI + objective metrics to evaluate it. Read the full post in Spanish here.

→ 10 Things Software Developers Should Learn about Learning

From CACM, 10 Things Software Developers Should Learn about Learning. The title says it all, but it’s also a good reminder for all of us. I especially loved item 6 when it comes to AI: “The Internet Has Not Made Learning Obsolete.”

→ AlgoMasterIO — Bite-sized DSA & LeetCode Patterns

If you’re brushing up on data structures and algorithms, the AlgoMasterIO playlists are worth watching. They break down core concepts into high-signal videos (arrays, graphs, linked lists, etc.) and focus on what actually matters: recognizing patterns rather than memorizing problems. The LeetCode patterns series is especially useful covering essentials like sliding window, two pointers, prefix sums, and more, which are key to solving a wide range of problems efficiently.

🔍 Resources for Teaching CS

→ What is tenure really?

This is a fascinating episode that elaborates on and digs into what tenure is from a legal perspective, as well as the arguments for why it is important beyond the classic reason of academic freedom. It’s a must-listen for anyone who wants to better understand the context and discuss the importance of this idea.

→ Guidelines for teaching informatics

The “Guidelines for Teaching Informatics: Practical Strategies for European Classrooms” have been officially published, providing practical approaches to make computing education more engaging, inclusive, and meaningful for all students; the document is part of a broader EU package on digital education (including AI, disinformation, digital content, and informatics) and will be translated into all 24 EU languages.

→ GenAI Resources for the CS Curriculum

This week in San Diego, the Workshop on Generative AI in Computer Science Education took place. I was browsing through the resources on their website, and they’re worth a look: they’ve published several reference courses that integrate GenAI across the CS curriculum (introductory programming, OOP, algorithms, software engineering, working with large codebases, etc.), with slides and projects that instructors can adapt to their own classes.

→ AI: Understanding the Game

I’d like to recommend a book in Spanish written by Margarita Padilla García, a computer engineer and educator who has spent decades involved in free-software activism. The book doesn’t try to provide definitive answers. Instead, it presents a collection of technical, historical, and philosophical ideas so that each reader can draw their own conclusions.
https://traficantes.net/sites/default/files/pdfs/TDS-UTIL_33_IA_web.pdf

→ National Academies Report — Data & Computing Competencies (K–12)

A new National Academies consensus report explores what it takes to develop data and computing competencies in K–12 education. It outlines seven core competencies and highlights a key challenge: while interest in these areas is growing, access and coherence across grades remain uneven. The report calls for coordinated efforts across curriculum, teacher preparation, assessment, and infrastructure to ensure all students can meaningfully engage with data and computing.

🦄 Quick bytes from the industry

→ Codex, Code Review, and Taste

What a fascinating conversation, especially when it touches on the intersection between research and engineering, and how each one depends on the other; choosing a company based on its consumer impact (including internal impact, like developer tools); how he still reviews the code even while heavily using Codex; how much code review should be done by an agent versus a human; and why deep technical skills are still valuable. I’ll definitely come back to this many times. Michael also has a really interesting technical career story.

I think we do a lot of very meaningful things (at Codex) but if the model weren’t very good it wouldn’t really matter what we did.
I really enjoyed working on Calendar because it was consumer-focused, and I shipped it to a lot of people. I went to Facebook because I thought it would be a huge consumer space, but it ended up not being consumer-focused at all. It turned into developer tools. My users were basically my friends at work. It was just about 20,000 people—not a billion—but it was good enough for me at the time. Then, thinking about OpenAI and the chance to come back to consumer—or at least have a large user base—was exciting. Now, working on Codex, I think there are over a million weekly users. I forget the exact number right now, but it just keeps growing. It’s more like a vertical line than even a hockey stick. That’s way more than the 20–40,000 developers I might have affected at Meta. I’m actually a much bigger user of the Codex app now than I expected to be.
I’m not someone who doesn’t look at the code. For projects that are true prototypes—throwaway projects—I’ll sometimes not look at the code at all. It’s very freeing, and I get why people are excited about it. But for the code that goes into Codex itself, I still need to look at it. That’s pretty important because it affects everyone else. Over time, you start to get a sense that the model will be able to handle certain changes. You end up getting a lot more done. Sometimes I even feel a little bad writing code by hand because I think, “If I had asked the model the right way, it could have done this.” It’s like when you start with something small—just changing three lines—and then 30 minutes later you realize you’ve gone much deeper. Of course, we all still like typing sometimes.
If you asked what percentage of my code is written by me versus generated by the model these days, it would probably be 80–90% model-generated. For things like debugging tests or CI issues, I’ll often ask the model for help—like adding print debugging. Every time I sit down to write something, I ask myself, “Should I write this?” And the answer is almost always no. There are some lower-level things, though. The Codex harness—the part I spend most of my time on—is written in Rust. That means we can do operating-system-specific things. I also spend a lot of time on sandboxing, which is what ensures the security and integrity of the system so the model can’t go outside the bounds you set. I tend to write more of that part by hand because I need to be really sure it’s correct and that our test coverage is solid. Sometimes I’ll seed the structure myself—lay the groundwork for the pieces I care most about—and then let the model fill in the rest. A lot of what I use the model for now is refactoring. For example, I might build up a big pull request that does too many things, and then ask the model to split it into reviewable-sized commits.
I like the approach where the agent does multiple rounds of review until it’s confident that the result is worth a human’s time to look at. But generally speaking, we still review it before it goes in. I think our agents still have an empty file, like everybody else’s does. Sometimes you still find gaps in knowledge where more context needs to be added back in. There are things we just haven’t documented yet—things that the agent doesn’t know but that I, as a human, still happen to know. Because of that, we do still catch issues. One interesting thing is that people are also using AI to write their pull request summaries now, and the summaries across the team have gotten way better. So when I go into a review, it’s already been reviewed by Codex, and there’s a summary that clearly explains both the what and the why behind the PR. That definitely helps us get through reviews faster, which is good because there’s a lot more code to review.
I actively try to go deeper through the layers and understand how things work. Many times I’ve seen other people do this, and now I can do it too. There are problems that other people were able to solve that I couldn’t solve before simply because I didn’t know there was cruft between two layers of the system. I still think it’s important to push yourself to go through the levels of abstraction and understand things at a deeper level. Of course, I think this will change over time. But right now, the questions you ask the agent still affect the quality of the result you get. If you’re not asking the right questions, you might not get the best engineering solution. As things progress, maybe that will become another layer that gets abstracted away. I think that’s probably where we’re heading. I just don’t know the time frame. Things do seem to be moving faster than we expect. In general, learning how to ask the right questions is really important. Even for myself, I haven’t fully figured out what that means for someone who’s just starting out. I’m fortunate that I have experience to fall back on—that’s where my sense of taste or intuition about what to ask comes from. But if you’re new, I’m not entirely sure yet what the best path is. And it’s also hard to say because we don’t fully know where everything is going.

→ Hiring Senior Engineers + behavioral interviews

Austen McDonald led mobile hiring at Meta and served on hundreds of hiring committees during his time there. Ryan Peterman spoke with him about what happens behind the scenes in senior+ engineering hiring and how to succeed in behavioral interviews.

Ryan has mentioned that behavioral interviews often feel less concrete than technical ones for engineers, so this conversation also explores subjectivity and bias in the process.

🌎 Computing Education Community

University of North Carolina Wilmington is hiring a tenure-track Assistant Professor in Computer Science (Software Engineering) within its College of Science and Engineering—applications are now open.
The University of Houston is seeking a Postdoctoral Fellow in Curriculum & Instruction focused on ML approaches to modeling human learning and language. The position involves working with interdisciplinary teams, analyzing learning and language data (including NLP and LLMs), mentoring graduate students, and potentially teaching.
For academics already experimenting with AI, Google has launched a new fellowship focused on tackling real institutional challenges, from assessment and policy to research workflows and how AI reshapes learning. Participants work on one concrete problem within their institution while joining a cohort exploring a broader question: what should responsible AI in education actually look like? Applications for the 2026 cohort are now open.
Hamilton College is hiring up to two visiting faculty (Instructor/Assistant Professor) in Computer Science for 2026–27. Open to all areas, the roles emphasize teaching across the curriculum, from intro to advanced courses, in a liberal arts setting with strong student engagement. One-year appointment, teaching load of five courses, with competitive salary (~$97K–$100K). Review starts April 1, 2026.
The IEEE-CS (with ACM, CRA, and CSAB) is gathering faculty input on a key question: how should high schools prepare students for undergraduate computer science? The survey explores topics like CS exposure, mathematical readiness, study skills, and the evolving role of AI in student preparation. Results will inform community recommendations and future curriculum discussions. Worth contributing if you teach intro CS.
Congrats to Charles H. Bennett and Gilles Brassard, recipients of the 2025 ACM A.M. Turing Award for laying the foundations of quantum information science. Their decades-long collaboration helped redefine information itself, shaping fields from cryptography to computational complexity and driving today’s momentum in quantum technologies.
A new NSF-funded project is recruiting faculty to join a multi-institutional effort on replicable computing education research focused on student teamwork (team formation, functioning, and interventions). Participants can join as Design Team members (co-developing the study; workshop at UVA + stipend) or Replication Team members (implementing it locally). A great opportunity to contribute to more robust CER across institutions. Applications open now (rolling from April 1).
Duke University’s ECE department is hiring a Professor of the Practice (teaching faculty) in areas including software, computer architecture, AI/ML, and HCI. The role focuses on teaching (2–2 load), requires a PhD, and emphasizes contribution to Duke’s teaching mission. Applications are open until late June.
A new study by Tommaso Carraro explores how faculty approach Society, Ethics, and the Profession (SEP) competencies in computing curricula. While frameworks like CS2023 emphasize these areas, they often remain part of a “hidden curriculum” with limited support for instructors. The survey (10–15 minutes) aims to capture how faculty integrate and develop these competencies in practice. Worth contributing if you teach computing.
Looking to make early CS courses more engaging? The BRIDGES workshop at UNC Charlotte introduces a toolkit for integrating real-world data, visualizations, and interactive projects into CS1/CS2, Data Structures, and Algorithms. With APIs for Java, C++, and Python, it’s a practical way to help students connect computing concepts to real problems.
CCSC Midsouth 2026 — Nashville (Apr 10–11)
Submissions are now open for the UK & Ireland Computing Education Research (UKICER) Conference. Deadlines span April–June (papers, workshops, doctoral consortium, posters), with the conference hosted at the University of Cambridge. A great forum connecting researchers and practitioners across Europe and beyond.
Virginia Tech’s Echolab is recruiting CS instructors for a 90-minute co-design workshop on improving student participation in live coding lectures. Open to anyone who regularly uses live coding (no design experience needed), with a $75 gift card for participants. A nice opportunity to reflect on and shape teaching practice in this space.
The Professors’ Open Source Software Experience (POSSE) workshop will take place at Red Hat HQ (Raleigh, NC), helping faculty integrate open source and team-based development into their courses. The program includes an online phase followed by a 2.5-day in-person workshop. No prior OSS experience required, with travel support available. Applications open (rolling until April 20).
The MAP-CS project is collecting faculty input on how institutions review and update undergraduate CS curricula, with a focus on supporting innovation and liberal arts contexts. The 10-minute survey explores current practices and challenges, aiming to inform better resources and share insights with the CS education community.
A full-time Senior Education Researcher position is open (remote), ideal for those interested in applying education research skills in new contexts. Applications are reviewed starting March 23 and close April 3 (or until filled).
CRA-E is running a national survey on the challenges faculty face in making computing courses digitally accessible, ahead of the April 2026 ADA Title II requirements. The goal is to identify barriers, reduce faculty burden, and inform practical resources (tools, guidelines, and best practices). A quick 5–10 minute survey—worth contributing if you teach in the U.S.
The CCSC Central Plains conference will take place at Drury University, with early registration open until March 27.
JetBrains’ KotlinConf will feature an Education Track focused on teaching Kotlin and adapting to changes in computing education. The track brings together educators and industry to explore teaching practices and materials, with insights that extend beyond Kotlin itself.

🤔 Thought(s) For You to Ponder…

What I find interesting about this article in Spanish (Omnes) is the underlying real ethical issue: AI companies (the ones developing these models) should design systems that do not exploit deep human attachment mechanisms. That’s their responsibility. They must make it clear that AI does NOT feel.

I love this segment of A Vivir. What I like most is that they’re still committed to going out to places and talking to people in person. It’s not the same conversation you have with someone over the phone as the one you have face to face and you can tell the difference on the radio, in podcasts, and in life.

Christian García Bello is always worth listening to. Here are a few notes, not necessarily verbatim, that I jotted down while listening to him on this podcast with Alex Sanz Vicente and Javier G. Recuenco:

Science has communicated incredibly well and enjoys a lot of prestige. The arts can do the same: tell compelling stories without lowering the bar.
The sophistication has to live in the work itself. Whatever you put on top of it should be simple and act as a guide: point to an entry point, give a bit of a user manual, let the viewer explore it, and promise that there’s a reward at the end.
We keep educating our taste and our eye the more we consume: we start to notice more twists, more nuance, and so on.
So many people understand Spain through its art: Picasso, Dalí…
When uncertainty enters the picture, the human element has to come in too.
The idea of the “deep gaze” is great as well: everyone naturally does it within their own field. There’s also this idea of “sensory knowledge”—the kind you gain firsthand and that’s hard to pass on, something that emerges naturally. And then there’s “sound judgment”: expert intuition, that sense of “this works”, having a good eye for it.
His criterion for choosing materials is how he can use them to tell stories. I loved the idea of the poetics of concrete.

Ferran Adrià talks about his relationship with Juan Mari Arzak (01:08:40): Setting egos aside to achieve a greater goal, creating a movement and a global brand to position Spain on the world stage so that many people can benefit from it, directly or indirectly.

I’ve started reading a recent Vatican document. It has 164 sections, so it will take me a few days to get through it. I’m particularly drawn to the intersection between Christian anthropology and technology, essentially the relationship between humans and machines. I’m planning to write about this in my newsletter next week, focusing on topics such as cognitive capacities; the distinction between enhancement and replacement; and the difference between suppressing or substituting what is human versus integrating technology in ways that bring human potential to its fullest expression. I’m also interested in exploring the risks: the opacity of automation in sensitive areas, dynamics of isolation, the weakening of critical thinking, the rise of a kind of digital “spiritual marketplace” without genuine community, and the emergence of technological substitutes for ultimate meaning. However, I don’t want to focus only on the risks; I’m equally interested in exploring possible solutions.

Manu de La-Chica from Soul College recorded a great episode of Ágoras with Dominican friar Adrien Candiard. It’s about 30 minutes long, and it absolutely flew by.

A few ideas that really stuck with me:

It’s a mistake to think we have to be saints in order to be loved by God. We can become saints because God has loved us.
Maturity means being free. Free to do what I want—what I believe is right—not to please other people, God, or my parents. To be adults, we have to choose our own lives and actions. God wants us to be adults. He wants us to be adults because He desires a relationship of friendship with us, not one of servitude—and that’s a big difference.
We have to weigh situations carefully. We need to act like adults, like people who recognize where the good is and move toward it. You can’t write a manual of Catholic morality that covers every situation, because morality, properly understood, is the science of the particular. In your personal situation, in every action you take, there are a thousand different factors at play in your mind—and only you can discern the good in that moment.
Of course the commandments help. Of course the experience of others, of the Church, helps. Ignoring that wisdom isn’t a good idea. But in the end, you are the one who acts. And God knows that—He gives us this freedom, which is difficult, demanding, and what ultimately makes us adults.
God’s will is better. Yes, of course it is—but it’s better because it aligns with what I most deeply desire.
The problem with the Pharisees is that they can’t accept that prostitutes and tax collectors are saved—that it would be impossible to be with them if heaven were filled with people like that: “I don’t want to go in.”
And that’s why they killed Jesus—not for any other reason. They killed Him because He opened the Kingdom of God to everyone. Accepting God’s goodness toward others is a major challenge, especially for those who see themselves as righteous, who haven’t committed serious sins. But you can’t love God if you don’t love His will.
The work of the Spirit in us is to make us children of God, as St. Paul says in the Letter to the Romans. The Spirit’s action makes us sons and daughters. And prayer is about welcoming that work. Saying a lot of “Our Fathers” can be helpful—but it can also become mechanical. And that doesn’t please God. If I’m just repeating prayers to feel like I’ve done something, it means I haven’t understood that the one acting in prayer is the Holy Spirit—and I need to let Him act.
This isn’t easy. Here too, the demands of the Gospel are high. To be children of God and accept that the primary work in my Christian life is done by the Holy Spirit—not by me—requires a certain humility.

This article by Marcos Rodríguez Vega, a cybersecurity and AI researcher at the University of La Laguna, does a great job explaining the implications of the data generated by our everyday digital activities. Individual data points aren’t all that significant on their own, it’s the analysis of big data and the combination of many small traces that really matter: probabilities, statistics... And that has consequences. It moves from analysis to actually leveraging behavior as a highly sought-after commodity in the digital marketplace. Marcos talks about companies that trade in this data, the importance of safeguarding security and privacy, and the need for net neutrality. Really interesting.

Everyone has something to teach us. We can learn from anyone. When we start becoming polarized, that’s when things go wrong. We should have a deep and consistent love for the truth, while staying clear about the fact that we don’t own that truth. This article by Tsh Oxenreider really made me reflect: our affection can’t depend so much on who we naturally like or dislike. The way we treat others will only be truly just if we make an effort to see people the way God sees them. In every group, there will always be people we naturally click with and others we don’t. But if I rely only on my own perspective, I’ll end up showing warmth and enthusiasm to the same people over and over, while consistently excluding others. True fairness, then, means treating others based on how God sees them, even if they’re not the nicest or easiest person for me. That’s what love is: willing the good of the other and recognizing the good in them.

The Commonplace

On Making Friends & Partial Solutions🫸🫷

On Friday I asked whether there was any interest in me importing some of my old essays from The Art of Simple, the blog I ran from 2007-2020. Enough people said yes for me to at least test out this little idea, so this is me, officially testing it out…

3 months ago · 41 likes · 9 comments · Tsh Oxenreider

Related to today’s topic: the problems that come from blind faith in data.

What do you do to rest and recharge? Here are a few things I try to practice:

Reading: having a fixed place and time for it, almost like a sacred block in my schedule.
Moments of silence and deep disconnection are essential to detox from overstimulation: prayer, the rosary, etc.
Walking in silence while listening to enriching podcasts.
Letting myself be bored from time to time to free up my mind. I like going out into nature for that.

📌 Research Corner

As Juho Leinonen and colleagues argue, current approaches to teaching programming mostly add AI on top of existing models but that may not be enough. Their work points toward a deeper shift: learning programming through natural language as part of a new paradigm. In parallel, their award-winning paper “Probing the Unknown” (Best Research Paper at AICSEPAR 2026) explores how students interact with AI-mediated problems at scale, offering early empirical insight into how this transformation is already unfolding.

I took a slower and more careful pass through the Information Gathering Toolkit from Omni Institute, created by Oskar Burger, Shon Reed, and Bianca Gonzalez-De La Rosa. It covers three main methods of data collection: (1) surveys, (2) key informant interviews, and (3) focus groups, and it wraps up with a section on both quantitative and qualitative data analysis.

On the ethics side, the document makes it clear from the start that any data collection effort should adhere to principles like voluntary participation, confidentiality, professional competence, and justice—all grounded in the Belmont Report.

For surveys, the guide walks through how to design them, what types of questions to use (closed-ended, like Likert scales, or open-ended), how to improve response rates, and how to administer them. It also warns about sampling bias and recommends random sampling whenever possible.

When it comes to interviews and focus groups, it emphasizes the importance of having a well-developed discussion guide, building rapport with participants, and separating the roles of facilitator and note-taker. It also offers practical tips for handling tricky situations, like dominant participants or questions that are met with silence.

Finally, in the data analysis section, it distinguishes between quantitative analysis (means, percentages, standard deviation) and qualitative analysis (coding, identifying themes, collaborative interpretation), and stresses the importance of not projecting your own biases onto the data.

I’d recommend it.

🪁 Leisure Line

Two weeks ago, I visited TXRX Labs at the East End Maker Hub in Houston, a makerspace with several workshops, including woodworking. Some of these spaces would be perfect for crafting things in 3D. The next step is to sign up for some classes.

I didn’t write the email last week because I was traveling in New Mexico for spring break. I’m still processing the experience… So much beauty and great company.

I just finished my bracket! I know Duke or Florida could beat UH, but I can’t bring myself to pick against them. I also think Michigan could beat Iowa State, but both are great teams. I’m confident UCLA will make it to the Elite Eight, beating UConn. I’d love to know what you all think in the comments!

📖📺🍿 Currently Reading, Watching, Listening

The documentary about how IntelliJ IDEA was made.

Solid episode from Arthur Brooks about what leisure actually looks like. Good stuff.

I’ve mentioned Twenty Thousand Hertz before, the Dallas Taylor podcast all about the world of sound. In this episode, they dive into the sonic universe of the Harry Potter audiobooks, and it’s a fascinating listen.

I’ve had the closing song from this newsletter stuck on repeat since Sunday morning. A great deep dive into Taracá with Ángel Carmona. The final segment about the AI-generated song is brilliant: empowering words is like bringing Pancho Villa together with Zacatecas. No matter how sophisticated models become or how well they mimic what we call “judgment,” they’ll always be missing lived experience. They’ll lack context. They won’t have any real-world street smarts.

Jessie Buckley was the queen of the night at the 2026 Oscars. I loved her speech and her take on female empowerment. Two lines in particular are still echoing in my mind: when she said to her husband, “I’d have 20,000 more babies with you,” and when she dedicated her award to “the beautiful chaos of a mother’s heart.” The line about the babies is clearly an exaggeration, but it points to an overflowing experience, that desire for an almost infinite multiplication of life, something that resonates with the idea of vitalism. The second speaks to motherhood as a kind of creative abundance. In the way she frames it, motherhood doesn’t compete with her identity as an artist; it expands it. She doesn’t romanticize the chaos either. She acknowledges it and lives it intensely, but she places professional success within an emotional ecosystem, recognizing her husband and parents, aware that she’s part of something larger. Her vision of female empowerment is rooted in belonging, seeing herself within a network, honoring those who came before and those who will come after, and valuing life and relationships beyond any cost benefit lens. She celebrates both abundance and chaos as the driving forces behind her art.

If you’ve been following me for a while, you already know I’m a big fan of Arthur Brooks. I’ve already pre-ordered his new book. It comes out at the end of this month.

💬 Quotable

We aren’t always free to change things, but we are always free to live through them in faith, hope, and love so that in every situation we grow humanly and spiritually.
Real freedom does not mean being ruled by one’s impulses from one moment to the next. Just the opposite. Being free means not being a slave to one’s moods; it means being guided in a course of action by the fundamental choices one has made, choices one does not repudiate in the face of new circumstances.

― Fr. Jacques Philippe (Fire & Light)

🌐 Cool things from around the internet

A collection of links to stuff I think are worth sharing.

🔗 Google Arts & Culture — curated historical and artistic content.

🔗 Archive of Archives — curated collection of digital archives.

🔗 gradient.horse — draw a horse and watch it run with the rest of the herd.

🔗 Visions of the Future — a graphic project by NASA. Beautiful.

🔗 The Tenth Muse (10M) — a new platform for discovering the world’s greatest art. Created by Paul Jun.

🔗 Series Graph — explore episodes through ratings graphs. The Game of Thrones episode rating chart is really cool.

Issue #36 of Computing Education Things was written while listening to:

🔗 Quick Links

🎧 Listen to Computing Education Things Podcast

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

Buy Me A Coffee

Taking the week off

Daniel Prol — Fri, 13 Mar 2026 12:02:35 GMT

Dear subscribers: I’m taking this week off for Spring Break to recharge ✨

Perhaps this weekend would be a good time for you to review the previous issues of the newsletter.

See you next Friday.

Thanks as always for your support 🙏❤️

#35 — In Defense of Software Craftsmanship

Daniel Prol — Fri, 06 Mar 2026 11:20:41 GMT

Reflections

Computing Education and Craftsmanship

Soetsu Yanagi is the writer who inspired this reflection. He always seems to enter my life at just the right moment.

In The Beauty of Everyday Things, he wrote:

With the exception of the hand of God, there is no creative tool comparable to the human hand. Its natural movements bring forth all kinds of beautiful things. No machine, no matter how powerful, can match that freedom of motion. The hand is the most precious gift nature has bestowed upon humanity. Without it, beauty would not exist.

Today, I want to offer a defense of craftsmanship in Computing Education and, more broadly, of the entire artisanal process in the moment we’re living through. A defense of the human right to participate in the process. The right to feel, to shape, and to transform the world of Computing Education with our own hands.

In Computing Education, craftsmanship emerges when an instructor designs an assignment, when a TA sits beside a student and traces the code line by line until the bug reveals itself…

In a world saturated with AI, craftsmanship stands as an act of resistance: an affirmation of creative freedom and of the joy that comes from shaping something patiently, from learning through trial and error, from refining with care.

If we allow AI to dictate the entirety of creative processes in Computing Education, we risk stripping educational work of its joyful dimension.

I have written before about the illusion of competence, the dangerous feeling of fluency when a solution flows but understanding remains hollow. AI accelerates fluency. It produces syntactically correct code, polished explanations, and elegant summaries. Craft, by contrast, slows things down. It introduces friction. In Computing Education, struggle is often where learning happens: when a student misplaces a brace, misunderstands scope, or confuses reference with value. Those moments are formative experiences as well.

Even if AI does make something for you and does it well, quality can drop fast if you’re not careful to polish and clean up everything afterwards. For code quality and product design quality. Always polish it yourself. That’s where craft lives. Paul Stamatiou’s 2025 Year in Review.

Reclaiming craftsmanship in the age of AI does not mean rejecting agents or LLMs. It means placing them properly. AI can help draft materials, generate variations, simulate scenarios, and expand access. But the final shaping of a learning experience, aligning it with educational intention, calibrating its difficulty, protecting the dignity of the learner, ensuring fairness in assessment, remains human work.

The relationship between craftsmanship, Computing Education, and AI should be one of complementarity. Each can expand the boundaries of what is possible. But only one can cultivate wisdom. In a world that seems obsessed with mechanical perfection, we need more than ever those hands that shape, that feel, that create.

Execution is great, but it isn’t enough

How much of what we do could AI actually do better? And how much of it is truly unique to who we are?

There is unease, uncertainty, and stress. Javier from the Tramontana Institute introduces the concept of digital intermediation:

Product designers do not design the journey itself, but the system for buying tickets; they do not stock the fridge, but set up the space where the products that will fill it are displayed and purchased; they do not create savings or investments, but design the screens where orders are placed to move money around. And that layer is now starting to fade.

In his view, AI agents are changing the rules of the game. We are paying for more machine hours than human hours. Should we reinvent ourselves? Just keep moving forward without knowing exactly where we are headed?

I am not sure. Let’s make sure the ball meets us in motion. And let’s not lose our taste or our judgment. Critique raises the bar. Navarre-born physicist and Prince of Asturias Award winner Pedro Miguel Echenique recently said that, in order to foster creativity, it is essential to trust people and promote collaboration.

Literature Mapping with Undergraduates

In this teaching practice byte (TPB), Dr. Brian Harrington discusses his SIGCSE Technical Symposium 2025 paper on Literature Mapping, a scaffolded, scalable, low-overhead way to introduce undergraduate students to research and bootstrap a student research group. They discuss how literature mapping helps students practice reading many papers in progressively more depth. His process assigns each paper to two different students, builds in flexibility for students who leave partway through, and culminates in a publishable artifact that students can be proud of. Moreover, he has found this helps him build a community that goes beyond the students graduating. Dr. Harrington has packaged up his process into a GitHub Repository and would love for anyone to adapt what he does to their own school.

Keeping reading groups consistent is always a challenge—there’s always something else that seems more “important” to do. But involving undergraduates seems really valuable to me, especially as an opportunity for them to build skills: something concrete for their CV, and also a way to explore whether the academic path is right for them.

I also love the community aspect. Building a community that can sustain and nourish itself over the long term is incredibly valuable.

🔍 Resources for Learning CS

→ How Researchers and Software Engineers I Respect Are Using AI Right Now

This week, I read several articles about how professionals I deeply respect are currently using AI in their work as researchers and software engineers. Yes, it’s another take on the topic but these are people who really know what they’re talking about. In case you’d like to explore further:

How I Use Claude Code

The Computist Journal

How I'm Using AI Today

3 months ago · 10 likes · 10 comments · Alejandro Piad Morffis

The Software Engineer Weekly

How I Use AI to Code

3 months ago · 23 likes · 3 comments · Kevin Naughton Jr.

→ COMPSCI 201 - Algorithms and Data Structures (Duke)

Owen Astrachan from Duke is teaching a DSA course this semester and is sharing everything online, including the lecture videos. What better plan for Spring Break than reviewing concepts like linked lists, sorting, and more?

→ Three Articles on Metaphors

Here are three articles centered on metaphors that I enjoyed this week:

The metaphors of artificial intelligence

The Counterfactual

What we talk about when we talk about LLMs

Faced with something new, people often try to understand it in terms of something they already know. There’s not much else we can do in these situations: we rely on metaphors and analogies and hope they lead us closer to something like the truth. But as George Lakoff and Mark Johnson pointed out in…

2 years ago · 19 likes · 16 comments · Sean Trott

The Absent-Minded Professor

The Metaphor is the Message

Thank you for being here. Please consider supporting my work by sharing my writing with a friend or taking out a paid subscription…

3 months ago · 18 likes · 4 comments · Josh Brake

→ Buy versus Build an LLM

The ACM Technology Policy Council has released a new TechBrief titled “Buy Versus Build an LLM,” examining the trade-offs governments face when deciding whether to develop large language models domestically or procure them externally, and proposing a strategic framework that considers factors such as sovereignty, safety, cost, capabilities, cultural fit, and long-term sustainability.

→ kuva

A scientific plotting library in rust, along with cli binary with the option to plot directly into the terminal. Here you can take a look.

→ #NoStupidQuestions Video Series

Have you ever had a data question that you were afraid to ask? Like, what’s the difference between data and statistics? Join ICPSR as they use cats, chocolate and more to unravel data mysteries. The #NoStupidQuestions series is available now.

🔍 Resources for Teaching CS

→ Introduction to Graph Algorithms (CMU in Qatar, Fall 2024)

At SIGCSE TS in St. Louis, I had the chance to meet Ryan Riley and dig a bit deeper into his teaching practice and I’m glad I did, because he has some excellent recorded lectures on YouTube, including a Graph Algorithms class where he opens with a memorable “architect’s view” of graphs: he models an apartment floor plan as a wireability graph (rooms as vertices, shared walls as edges), then adds edge weights as “wire cost” to motivate core ideas like paths, degree, connectivity, and classic optimization problems like the minimum spanning tree (Kruskal/Prim) and shortest paths (Dijkstra), linking the abstractions back to real-world systems like electrical wiring and navigation tools such as Google Maps/Waze.

→ Online Resources from Mark Liffiton

Mark Liffiton is teaching three courses this spring, and he offers a variety of valuable online resources on his personal website:

→ AP CSA Java 2025+

CSAwesome2 is the updated College Board–endorsed AP Computer Science A (Java) curriculum aligned with the 2025–2026 revisions, offering interactive Runestone-based exercises and teacher resources.

→ Reimagining Grading

I stopped by the Pensive AI booth at SIGCSE TS, but honestly I didn’t pay much attention at the time. I’ve since checked out the demo, though, and it looks impressive. It also seems like they’re doing well, at least judging by their seed funding.

→ Teaching Advanced Algorithms at the High School Level

Justin Skycak wrote this freely available textbook to support Math Academy’s former Eurisko program (2020–23), which scaffolded high school students to master’s/PhD-level AI coursework, including reproducing research papers in Python, with plans to bring the full curriculum to the Math Academy platform.

→ SQL Crack

Before teaching students how to tune a query, they first need to understand how it works. SQL Crack visualizes joins, filters, CTEs, and aggregations as a clickable flow graph, with column-level lineage and built-in performance hints. A useful tool for database instructors working with real-world SQL examples.

→ “CS and SE Education, post-AI” by Titus Winters (Adobe)

This was the Saturday keynote talk given by Titus Winters. I saw people actually stand up and applaud when it ended. I really think it’s a talk that everyone in CS departments should watch.

You can also find the rest of the SIGCSE TS 2026 videos in this YouTube playlist.

🦄 Quick bytes from the industry

→ All the Claw things

Peter Steinberger (now part of OpenAI), the Austrian behind OpenClaw, in an interview with Lex Fridman:

→ VP vs CTO roles and responsibilities

Bryan Cantrill compares the roles of VP of Engineering and CTO, focusing on inward mechanics versus outward customer focus. Go to the 48:03 mark.

→ Breaking Into Data Science in 2026: What Actually Works Now

The entry-level data science market has shifted. In this 11-minute breakdown, a senior data scientist who’s worked at Stripe, Uber, and a startup explains what’s non-negotiable in 2026, how LLMs are reshaping the skill stack, and why starting as an analyst may be the most pragmatic way in.

🌎 Computing Education Community

Aalto University’s Department of Computer Science is hiring three full-time, permanent Lecturers (Bioinformatics/Health Data Science, General CS, and Software Engineering), with applications due March 16, 2026.
The Learning & Education team at OpenAI is hiring for a Strategy & Ops (Education) role.
Artturi Tilanterä recently graduated as a Doctor of Science (Technology) from Aalto University, where he investigated students’ difficulties in learning computer science (particularly basic algorithms) and his dissertation, featuring a self-drawn cover illustration, explores misconceptions and foundational challenges relevant to those teaching programming, data structures, algorithms, and conducting research in learning analytics or educational data mining.
Alex Chao (UC San Diego) is inviting graduate students to join a new online community emerging from SIGCSE TS, which plans to host reading groups, paper review sessions, research discussions, panels, and social events.
The Raspberry Pi Foundation’s research seminar series continues on 17 March (17:00–18:30 GMT) with Rebecca Fiebrink (University of the Arts London) speaking about “Teaching AI to creators,” exploring how AI can support creative practice across audiences from children and students to professional artists and how it can enable new forms of creative expression beyond text-to-media systems. Sign up here.
Mats Daniels (Uppsala University) shared a postdoctoral opening in the NordRAI project (Nordic Education at the Frontline of Responsible AI), a NordForsk collaboration with Aarhus University, University of Eastern Finland, and Western Norway Research Institute; the position focuses on the Swedish context with broader Nordic collaboration, and the application deadline is April 7.
Anna Ritz is organizing the Computational Biology Peer Mentor Network (CompBioPMN), a three-year program bringing together ~20 researchers and faculty at primarily undergraduate institutions to build community, reduce isolation, and support career development in computational biology and bioinformatics; late-stage graduate students and postdocs interested in teaching-focused institutions are also encouraged to apply (deadline: May 1, 2026).
El Camino College is hiring a full-time, tenure-track Computer Science faculty member to teach courses such as assembly language, data structures, Linux, C++, Java, Python, MATLAB, and data science, with responsibilities that include curriculum development, student engagement, and participation in departmental and campus activities (application deadline: April 13, 2026).
A little self-promotion here: We’re announcing a new Artificial Intelligence track in the M.S. in Computer Science program at the University of Houston (Katy Campus), launching Fall 2026. Join one of our information sessions next week to learn more.

🤔 Thought(s) For You to Ponder…

Maca Gil mentioned on this podcast that they’ve been trying to use ChatGPT to create storyboards for animated films, but they haven’t been able to make it work. She also brought up this article about Pocoyó, which claimed the series was going to be produced entirely with AI. In the end, though, they had to walk that back because they couldn’t get the results they were hoping for. According to one of the illustrators on the podcast, AI still doesn’t have enough creativity.

Along the same lines, Daniel Cartagena, Spanish translator for The Simpsons, said on A vivir:

If there’s any respect for creativity, for the audience and for the translator, AI shouldn’t be used. The creativity a voice actor brings, the soul they pour into it—that’s not something AI can replicate. And I don’t think it’s worth it just to preserve the original actor’s voice. Our actors bring so much to the table.

On the relationship between money and evangelization. A really interesting and well-explained article by Javier García Herrería. I’ve always been happy to pay when I see that the content offers real added value. Here in the U.S., it’s totally normal (Edward Sri, Katie Prejean McGrady, Matt Fradd, Christopher West…). Hopefully it becomes more common in Spain. It just makes perfect sense.

Interesting framework proposed by Josh Brake arguing that genAI should be evaluated based on whether a task’s value lies in the product or the process, and on the role of the human element as instrumental or intrinsic.

I’m a big fan of Cal Newport and his books. He’s the kind of CS educator I want to become (even though he’s more of a pure CS guy haha). He just launched his “masterclass”!

I’m glad Computer Science has its own category in these prestigious awards.

The fact that the new MacBook Neo uses an iPhone chip brings some real technical challenges (thermal management, memory/buses, I/O/OS, performance…).

Very interesting: Brands now need a full editorial team, almost like a newsroom, now that algorithms reward consistency and saturation.

📌 Research Corner

Oliur has been using these AI tools to build. When it comes to terminals, I love using Warp. It’s easy to use and read. Whenever I just want to write plain text in a terminal, I use it. I tried Agentation in my project, and even though it wasn’t the right fit, the concept is really cool: you can annotate elements directly in your UI, and Claude Code picks them up in real time. If you’re building with React, it’s absolutely worth exploring.

This week we kicked off the ITiCSE WG10 meetings. We’ve got faculty from Boston College, Franciscan University of Steubenville, Florida International University, the University of Münster, the University of Illinois Chicago, Falmouth University, the University of Edinburgh, Monash University, the University of St. Thomas, Furman University, Berea College, the University of Illinois Urbana-Champaign, and the University of Houston. Only part of the group was in this session. We have colleagues joining from other parts of the world, but because of the time differences, we’ve had to split into two meetings. This WG is going to be quite an adventure, and it’s truly an honor for me to be part of it with this group from day one.

Related to the topic of my WG, this week I heard Adolfo Berraquero talking with Adriana Carvajal:

The faculty needs to dedicate quality time to teamwork, and that requires generosity from both sides:

In a world where technology is advancing so fast—and will keep advancing—everything is going to get crazier and crazier, and in the blink of an eye we’ll be seeing unbelievable things. I think human values are what will really make the difference. Building meaningful relationships, being generous, truly caring, and being honest about what I promise and what I deliver. In the end, what will really make people say “wow” is the person behind it—that’s my opinion.

The “sense” required to distinguish code that works from code that is architecturally sound. If coding is solved, the competitive advantage becomes your own taste—the true differentiator:

On the other hand, when it comes to skills, Adolfo says he sees something positive in the empowerment people feel now: the idea that you can build a website in two hours using just an audio prompt, without knowing how to read a single line of code, and still create a functional platform. But he also says that doesn’t guarantee you’ll be able to make a living from that platform or website. Because if you can do it, anyone can do it.

Very closely related to my current project: Hannah Walden argues that in the age of vibe coding and agentic AI tools, block-based programming may be more important than ever, because as syntax becomes less of a barrier, the real focus of computer science education shifts to problem solving, logical reasoning, debugging, and system design: areas where blocks excel by making structure and logic visible, reducing unnecessary friction, and supporting scaffolded learning alongside AI rather than hiding thinking behind generated code.

New from CRA-E: At SIGCSE, CRA’s Education Committee released a white paper titled Sowing the Seeds: Preparing Students in Research Computing Departments for Teaching-Focused Careers, outlining how departments can better prepare PhD students for teaching-centered academic roles and strengthen instructional capacity across computing programs. This builds on earlier CRA guidance, such as Laying a Foundation: Best Practices for Engaging Teaching Faculty in Research Computing Departments, which details how to equitably support, evaluate, and advance teaching faculty. As shared previously, the CRA Taulbee Survey also reports current salary data for teaching faculty.

This paper tested six types of compiler error messages in an online introductory programming course, including two generated by GPT. Students receiving GPT explanations repeated the same error 23% less often and resolved issues with 35% fewer attempts. Interestingly, long generic explanations performed worst, often misleading learners when they didn’t match the exact error. Over time, however, none of the message types significantly changed how often students made mistakes.

In my Numerical Analysis class this week, we discussed optimization and gradient-based methods. After the lecture, I came across this very clear visual introduction to PyTorch that explains tensors, automatic differentiation, and gradient descent in a simple and intuitive way. If you like visual explanations of ML concepts, this might be a useful resource.

I was reading this information gathering toolkit this week. It’s great. It’s available in both English and Spanish.

🪁 Leisure Line

Been waiting for my docker edition of Communications of the ACM :) I completely agree with Moshe Y. Vardi’s insight in this issue: at conferences, we should discuss computing research and that’s it.

Every other Friday I go to College Station, TX. Last time I took this shot at sunset with my iPhone 16. I love these moments.

📖📺🍿 Currently Reading, Watching, Listening

I highly recommend Frankenstein by Guillermo del Toro. I absolutely loved it. For me, it’s one of his best films so far. It’s truly a masterpiece: visually stunning, and it has a bit of everything: darkness, romance, gloom, transcendence, gore, hope, light… The performances are outstanding, too. It’s also a fantastic starting point for conversations about bioethics, identity, and other moral dilemmas like whether living forever is actually something we should want.

I loved this Song Exploder episode featuring the song ‘Como un Pájaro’ by Silvana Estrada.

I haven’t listened to live radio in years (just on-demand podcasts) but I really enjoyed this show by Sara Cox on BBC Radio 2. What makes it special is that real sense of live companionship. I listened to it on demand, of course, and loved it, but it airs every Friday in case you want to catch it live. There are some really heartwarming calls from listeners, like kids calling in after school just to say hi or send a kiss to their grandma. It’s pure humanity.

When I lived in Colombia in 2016, I was working at Platzi in Bogotá. I got to experience up close the peace referendum with the FARC and the government of Iván Duque. Álvaro Uribe Vélez, from the same political party as President Duque, has been a central figure in Colombia’s recent history. If you’re interested in everything that’s happened in the country politically and socially, I recommend listening to this episode. Uribe talks about his trajectory, his presidency, whether there’s any scenario in which he wouldn’t live in Colombia, and much more. Andrés from Atemporal ended up producing a really solid historical document with this one.

I discovered that the Witness History podcast from BBC World Service is experimenting with generative AI in some of its episodes. The first adapted episode, “The World’s First Labradoodle,” is already available on YouTube. It seems like a thoughtful use of AI respecting the original storytelling while trying to expand the reach of an audio format to new visual audiences. The visual work was created by the studio 1UpMedia.

Yesterday I watched “Broken Mary: The Kevin Matthews Story” on Credo+. I rented it for €3 with their 30% Lenten discount. Kevin Matthews is a legendary radio host who, after being diagnosed with a serious illness and watching his life fall apart, comes across a broken statue of the Virgin Mary and everything suddenly changes. Kevin discovers his true calling: he becomes Mary’s “roadie,” traveling with the statue and bringing it to different places, offering his listeners something deeper than entertainment: hope. Very inspiring!

🌐 Cool things from around the internet

A collection of links to stuff I think are worth sharing.

🔗 Agentation — point at bugs, let AI fix them.

🔗 Jordan Rosenberg Portfolio — a truly unique-looking portfolio.

🔗 Riffo — AI-based file naming for Mac.

🔗 Sandbar — a ring that can take voice notes and control your music.

🔗 This McDonald’s Ramadan campaign is brilliant.

🔗 I hadn’t seen this NYT post comparing the performances of the gold and silver medalists in figure skating. Hypnotic.

🔗 This Substack in Spanish is a powerful visual space for learning and reflection.

🔗 WalkmanLand pays tribute to the pocket tape devices of yesterdays: the Walkmans.

🔗 COiN — Cutter That’s As Small As a Quarter.

🔗 Hotel Retro — Vintage Luggage Labels from Tokyo to Buenos Aires.

Issue #35 of Computing Education Things was written while listening to:

🔗 Quick Links

🎧 Listen to Computing Education Things Podcast

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

Buy Me A Coffee

#34 — ITiCSE-WGR '25 - A Deep Dive

Daniel Prol — Fri, 27 Feb 2026 11:20:39 GMT

Reflections

ITiCSE-WGR 2025: Working Group Reports

Nijmegen city center, host of ITiCSE 2025

I’m excited to share that I’ve been selected as a member of ITiCSE 2026 Working Group 10 (WG10), titled “Teamwork in Computing Education: Skills, Values, and Virtues.” I’m really looking forward to meeting such incredible people whose work I deeply admire for its quality and impact. Ready to learn a ton. Can’t wait!

After hearing the news, I decided to dive into last year’s WG reports, which were recently published in the proceedings. As I read through them, I didn’t just see nine major research efforts but a set of signals about where computing education research is headed. What we choose to measure. What we’re concerned about. What we’re trying to rethink and reimagine.

What follows isn’t a paper-by-paper summary. It’s a structured synthesis of the patterns that, in my view, emerge across all nine reports, grounded in each group’s contribution.

Expanding what counts as knowledge

At least three reports challenge narrow views of knowledge in computing education:

CARD (Cards for Alternative Research Design) encourages CSEd researchers to reconsider paradigms and methodologies. It argues that methodological fluency (not just methodological familiarity) is essential for the field.
Notional machines in Data Systems expands the concept beyond programming into databases and data systems, highlighting variation and under-specification in conceptual models.
AI concept inventory for Non-Experts (AICI-NE) formalizes what AI literacy means by identifying key concepts and common misconceptions.

Signal #1:
The field is shifting from teaching content to making knowledge structures explicit: conceptual models, misconceptions, research paradigms, and epistemic scaffolds.

Measurement is maturing

Four reports demonstrate a clear shift toward rigorous evaluation:

Capstone Assessment Study uses surveys and interviews across institutions to examine fairness and contribution recognition.
Two-Stage Exams (TSEs) combine quantitative grade data with qualitative perception data across institutions.
Fairness in Student Group Formation integrates a systematic review, tool analysis, educator interviews, and student focus groups.
Broadening Participation through Undergraduate Research employs mixed methods, including statistical analyses of identity and access.

Signal #2:
The field is no longer satisfied with “we tried this in one class.” It is increasingly multi-institutional, mixed-method, and evidence-driven. This shift mirrors broader trends in the learning sciences and educational measurement.

Fairness is no longer peripheral

Fairness appears across multiple reports, not as a footnote, but as a central design concern:

Group Formation Fairness explicitly interrogates definitions of fairness and algorithmic allocation mechanisms.
Capstone Assessment reveals tensions between instructor trust in peer evaluation and student perceptions of risk.
Undergraduate Research Participation highlights differences in scientific identity and access among marginalized students.
GenAI Ethics in Higher Computing Education foregrounds equity, bias, data provenance, and institutional disparities.

Signal #3:
Fairness is shifting from an “ethical add-on” to a design constraint (e.g., group allocation algorithms, assessment weightings, access to AI tools). These are no longer neutral technical decisions; they are socio-technical systems.

Assessment is being reimagined

Assessment is one of the strongest cross-cutting themes:

Two-Stage Exams show positive student perceptions and performance effects across contexts.
Capstone Projects reveal the difficulty of evaluating teamwork, hidden labor, and stakeholder roles.
AI Concept Inventory attempts to build standardized instruments to assess non-expert understanding of AI.

Signal #4:
We are moving from “How do we grade this?” to “What exactly are we trying to measure, and what does that measurement do to students?” Assessment is becoming more pedagogically grounded.

Generative AI: beyond panic

Two reports directly engage with GenAI:

Ethical & Societal Impacts of GenAI develops the ESI-Framework for institutional decision-making.
The Rest of the Robots focuses on GenAI in post-introductory courses and instructor practices.

Notably:

The conversation has moved beyond “ban or allow.”
The focus is on integration, policy alignment, curriculum design, and workforce expectations.
Ethical, social, and academic integrity concerns are embedded within pedagogy.

Signal #5:
GenAI is being normalized as infrastructure. The real question is not whether to adopt it, but how to integrate it responsibly and pedagogically.

Participation and Identity

The Undergraduate Research working group frames research exposure as a mechanism for broadening participation. This complements ongoing work on fairness in group formation and assessment.

Signal #6:
Belonging and identity are no longer treated as “diversity side topics.” Instead, they are increasingly understood as structural design outcomes of educational programs.

Final Reflection

If I had to summarize ITiCSE-WGR ’25 in one sentence:

The field is shifting from experimenting with ideas to engineering reliable, fair, and conceptually grounded systems for computing education.

And that shift matters.

Because once a field begins building infrastructure (measurement frameworks, fairness models, validated instruments…) it is no longer operating in an exploratory mode.

That is good news.

I’d love to hear your thoughts in the comments.

The 2026 Tech Job Market: A Tesla Recruiter’s Perspective

Ben Chen explains how he stumbled into recruiting during the pandemic, realizing he had essentially been doing it his entire college life through leadership roles and community-building. After starting in a hyper-competitive agency environment that almost pushed him out of the profession, he transitioned in-house to gain ownership and real business impact. A cold DM from a Tesla recruiter changed everything: the pitch of building teams for real-world robotics and autonomous systems felt surreal. He joined Tesla and spent about 2.5 years helping build teams across Autopilot and Optimus describing it as building a startup inside a big company, creating functions from scratch while contributing to world-changing technology.

Ben’s view of the market is direct: it’s tough for juniors, but not hopeless. Companies increasingly prefer hiring one senior engineer over many juniors, especially now that AI tools amplify senior output. The traditional ladder (join entry-level and slowly climb internally) is less reliable than it used to be. Instead, juniors may need to seek high-growth startups where they can compress years of learning into high-intensity cycles of ownership and responsibility. He frames hiring as probabilistic rather than deterministic: interviews aren’t perfectly fair, internal dynamics shift, and headcount can disappear overnight but strong candidates increase their odds through preparation, real impact, and enough attempts. In his words, if you can’t get a job in 2026, it’s probably your fault, not as an insult, but as a reminder that control still exists. The market rewards merit, and world-class candidates will not stay unemployed.

Universities, he argues, are naturally 5–8 years behind industry. A CS degree provides opportunity, but it won’t carry you to the finish line. CS students must differentiate themselves through meaningful open-source ownership, novel research, founding projects, or building real products. He calls GPA “a huge scam” because it rewards one-shot performance and discourages risk-taking; unless it’s very strong (around 3.7+), it may be better omitted. Resume strategy should prioritize undeniable impact over buzzwords, since keyword stuffing often hurts more than it helps. The ATS exists, but most rejections involve human screening supported by keyword filtering (not a “magical” AI auto-rejection system!).

Recruiting is as much psychology as it is technical evaluation. Candidates must prove both that they can do the job and that they genuinely want to be at that specific company. Real interest requires deep research into a company’s edge, market position, and strategy. In an era of AI-assisted resumes and interviews, referrals and authentic networks matter more; the best networking strategy is building real relationships early and creating impact together. At the offer stage, entitlement or excessive demands can derail everything. Negotiation power doesn’t come from hacks, it comes from value/impact. If you are genuinely high-impact, companies will move for you.

🔍 Resources for Learning CS

→ What Civil Engineering and CS have in common

I recently read a BLOG@CACM post in which a Purdue University computer science professor reflects on the lessons he learned from his father, a retired civil engineer who served in India’s federal telecommunications department. Although their professional worlds seem vastly different, he shows how principles such as building for longevity, testing rigorously, and maintaining unwavering integrity translate powerfully into a meaningful career in computer science.

With the rise of AI, however, I worry that we may slowly stop respecting these principles.

→ The Missing Semester of Your CS Education (IAP 2026)

The Missing Semester of Your CS Education (IAP 2026) is a practical course that teaches students to master essential computing tools from the command line and version control to debugging, packaging, and AI-enhanced workflows so they can work more efficiently and effectively throughout their CS careers.

🔍 Resources for Teaching CS

→ Mathematics everywhere – not just in the classroom

MathCityMap is a platform that enables out-of-class mathematics activities. I like the fact that it combines the idea of math trails with the possibilities offered by smartphones.

→ Academic Leadership

Academia doesn’t train you to be a leader. In this episode, Jen Heemstra (WashU) talks about her book Labwork to Leadership, which aims to help fill that gap.

→ Effectively Teaching Learners with ADHD

In Episode 14 of Designed for Learning, host Jim Lang and guest Karen Costa explore how educators can effectively support students with ADHD by shifting from a deficit-based mindset to one that recognizes the ADHD neurotype as an “open house” with unique strengths and challenges. Costa argues that the traditional medical model focuses too much on deficits, whereas the ADHD brain often exhibits remarkable creativity, divergent thinking, and the capacity for “hyperfocus” (a deep flow state triggered by interest and passion). While these students frequently face executive function hurdles and a profound sense of shame when they fall behind, Costa recommends using flexible structures, such as grace periods and “think like a tree” adaptability, to keep them engaged. To mitigate challenges, she suggests externalizing information through multimodal reminders, checklists, and proactive outreach, emphasizing that these inclusive strategies benefit the entire classroom, not just neurodivergent learners.

→ BRIDGES

BRIDGES is an NSF-funded project that provides ready-to-use interfaces and visualizations (in C++, Java, and Python) built around real-world datasets such as IMDB, Wikidata, and USGS. It helps integrate data structures and algorithms into introductory CS courses without changing the core content, while adding applied and visual engagement. Thanks to Kalpathi Subramanian, Erik Saule, Paula Gookasian, and Jamie Payton for creating it.

→ Book of paper and pencil Python exercises

A GitHub project offering pencil-and-paper puzzles to teach basic Python syntax without a computer, designed for introductory-level students. The author, Peter Drake, piloted it in an introductory data science course and reports positive student reception. You can find more public course materials by Peter here.

🦄 Quick bytes from the industry

→ Boris Cherny on the Lenny’s Podcast

Lots of great insights from Boris Cherny (Anthropic) in his interview with Lenny Rachitsky about how AI is transforming the role of the software engineer.

Click the link and read the full summary.

Here are some of the most interesting quotes from the episode (I don’t necessarily agree; I’m just mentioning them):

I still do a lot of coding. A hundred percent of my code is written by Claude Code. I have not edited a single line by hand since November. Every day I ship 10, 20, 30 pull requests. I do look at the code. I don’t think we’re at the point where you can be totally hands-off, especially when there’s a lot of people running the program. You have to make sure that it’s correct. You have to make sure it’s safe and so on. And then we also have Claude doing automatic code review for everything. Here at Anthropic, Claude reviews 100% of pull requests. There’s still a layer of human review after it, but you still do want some of these checkpoints. You still want a human looking at the code, unless it’s pure prototype code that it’s not going to run anywhere.
Claude is starting to come up with ideas. It’s starting to get a little more like a co-worker. Looking through feedback, it’s looking at bug reports, it’s looking at telemetry for bug fixes and things to ship.
I think at this point it’s safe to say that coding is largely solved. At least for the kinds of programming that I do, it’s just a solved problem because Claude can do it.
There is a beauty in programming, especially in functional programming. There’s also a beauty in type systems. There’s a certain kind of buzz you feel when you solve a really complicated math problem—it’s similar to the feeling you get when you perfectly balance the types or when a program comes together in a truly elegant way. But that’s not the whole story.
There’s always space to enjoy the art and to kind of do things by hand if you want. Some Anthropic engineers still write C++ by hand on weekends.
I think for people that are using Claude code, that are using agents to code today, you still have to understand the layer under. But yeah, in a year or two, it’s not going to matter.
I don’t have to do the tedious work of coding anymore, because this has always been sort of the detail of it. It’s always been the tedious part of it and kind of like messing with a git and kind of using all these different tools. That was not the fun part. The fun part is figuring out what to build and coming up with this. It’s talking to users. It’s thinking about these big systems. It’s thinking about the future. It’s cooperating with other people on the team. And that’s what I get to do more of now.
If you know that it works correctly and efficiently, then you don’t actually have to know all the details.
I have never enjoyed coding as much as I do today, because I don’t have to deal with all the minutia. So for me personally, it’s been quite exciting. This is something that we hear from a lot of customers, where they love Claude Code, because it just makes coding delightful again.
Try to be a generalist, more than you have in the past. For example, in school, a lot of people that study CS, they learn to code, and they don’t really learn much else. Maybe they learn a little bit of systems architecture or something like this. But some of the most effective engineers that I work with every day, and some of the most effective, like product managers, and so on, they cross over disciplines.
On the Claude Code team, everyone codes: product manager codes, engineering manager codes, designer codes, finance codes. Some people have specialties. For example, I code a little bit more versus our PM does a little bit more coordination or planning or forecasting or things like this.
The people that will be rewarded the most over the next few years, they won’t just be AI native, and they don’t just know how to use these tools really well, but also they’re curious, and they’re generalists, and they cross over multiple disciplines and can think about the broader problem they’re solving, rather than just the engineering part of it. Some of the strongest engineers are hybrid product and infrastructure engineers, or product engineers with really great design sense or an engineer that has a really good sense of the business.
I think by the end of the year everyone’s going to be a product manager and everyone codes. The title software engineer is going to start to go away. It’s just going to be replaced by builder.
Programming has always changed.
There’s always this question, should I learn to code? In a year or two it’s not going to matter.
With traditional engineering work, if you’re working on product, you want to understand the infrastructure, the runtime, the virtual machine, the language, whatever that is, the system that you’re building on. If you’re working in AI, you just really have to understand the model to some degree to do good work.

→ Owning the AI Pareto Frontier

A very interesting (although highly technical) discussion on this recent episode of Latent Space. Great listen. Jeff Dean is a legend.

Last semester in the AI course with Jianyi Yang, I really got to see just how deep this world can get. I either didn’t know or had only a surface-level understanding of many of these ideas, and a lot of them were completely new to me. That’s why I really appreciated hearing someone like Jeff break it all down so clearly in this episode.

It really reminded me of what we mentioned about Andrej. He perfectly blends the ability to explain complex concepts with real industry experience. And kudos to the hosts as well for steering such a technical conversation so smoothly. Definitely worth the listen.

🌎 Computing Education Community

The Department of Computer and Data Sciences at Case Western Reserve University invites applications for a non-tenure-track faculty position (Assistant or Associate Professor) focused on teaching and student advising; learn more and apply here.
Dennis Reidy, a public health professor at Georgia State University, invites students to complete a 15-minute online survey examining how different academic disciplines relate to health and well-being. The goal of the survey is to inform future population health strategies.
Don’t miss the free ACM TechTalk on March 11 at 12 PM ET, “A Practical Introduction to Agentic Coding” with Microsoft Python/AI expert Marlene Mhangami, featuring production-ready patterns for building semi-autonomous AI agents with GitHub Copilot, MCP, and Python.
The NSF CyberAI SFS team invites the SIGCSE community to attend a virtual webinar today (11:00 a.m.–12:00 p.m. EST) to learn about the new NSF 26-503 solicitation supporting AI and cybersecurity education through a Scholarship Track and an Innovation Track.
SIGCSE Reads 2027 selections have been announced.
ITiCSE 2026 (Madrid, July 10–15) is seeking participants for two working groups: WG6 on Globalizing Computing Terminology and WG4 on Replication Studies in Computing Education Research.
DataEd 2026, a workshop bridging Computing Education and Data Systems research, will be held March 24 in Tampere, Finland (with keynotes by Craig Zilles and Azza Abouzied), and is offering two €1000 travel grants for interested researchers, educators, or students. Apply by March 2 AoE.

🤔 Thought(s) For You to Ponder…

When I read about other industries, I try to draw parallels with computing education. The creative process of a music producer like Raül Refree blends technique, time, patience, empathy, intuition, instinct, trial and error, and experimentation (more than just commercial calculation!). Computing education trains very similar processes: iteration, debugging, structured thinking, prototyping, reviewing, and teamwork. From that perspective, optimizing only for “the job market” doesn’t really seem like the most important thing, does it?

Last Sunday, the Winter Olympics came to an end. Pope Leo wrote a letter titled “Life in Abundance,” in which he highlighted the great value of sport.

At SIGCSE TS, I had the chance to learn more about the work behind the CRA Taulbee Survey thanks to Noah Cowit. Since 1974, it has been the primary source of data on enrollment, Ph.D. production, employment, faculty salaries, and demographic trends in CS, CE, and Information programs across North America. It’s an important resource for understanding the state of the field.

Building on what Boris from Anthropic was saying earlier, Marc Andreessen also believes that we need to understand every layer of the stack, that we’re more productive when we truly understand how the machine works. He argues that AI will generate a lot of code, but if we want to be truly great, we need depth. At the same time, he says the future of software engineering is less about typing and more about systems architecture, evaluation, and being a multidisciplinary builder. Interestingly, Boris used the word “builder” as well.

This reflection from the podcast Let Love about protecting time to let your thoughts move really got me thinking:

In our culture, there’s a temptation to just run, act, and put thinking on standby. We need to protect and make time for thinking today because there’s so much going on that we almost stop thinking in certain ways. Yet thinking is one of the greatest capacities of our human nature: our rational soul’s ability to ponder, reflect, and reason. If we don’t intentionally make time for that or become aware of our thoughts, we can easily rush through our days acting, doing, and reacting. Then we get to the end of the day and wonder, “How did I actually think today?”

Phil Eaton is starting a software research company.

For my friends who lean more toward the theoretical side of computer science, this Leslie Lamport episode is an absolute masterclass.

📌 Research Corner

Two books that are shaping the direction of my current research:

🪁 Leisure Line

I also used my trip to St. Louis to catch up with friends who are living there. Pipe and Marco, you guys are awesome. Until next time!

📖📺🍿 Currently Reading, Watching, Listening

This blew my mind this week:

🌐 Cool things from around the internet

🔗 Little Language Lessons — mini Gemini experiments for learning languages.

🔗 Tldraw — a whiteboard-style canvas.

🔗 granola.ai — enhance your notes in real time.

Issue #34 of Computing Education Things was written while listening to:

🔗 Quick Links

🎧 Listen to Computing Education Things Podcast

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

Buy Me A Coffee

#33 — Highlights from SIGCSE TS 2026

Daniel Prol — Sat, 21 Feb 2026 21:08:04 GMT

Reflections

This week I attended SIGCSE TS 2026 in St. Louis, one of the premier conferences in CER. It’s been a whirlwind of papers, panels, workshops, lightning talks, posters, keynotes, and countless hallway conversations. The program has grown so large that it’s impossible to attend every session or keep up with every contribution. I spent most of my time talking with people, but whenever I could, I dropped into sessions that looked especially interesting. Below are some sessions that caught my attention. For each one, I’ll briefly explain why I think it’s relevant and worth your time. Of course, there’s much more in the program, and I’m sure there are many other hidden gems waiting to be discovered. If you want to explore the full collection, I highly recommend browsing the proceedings. As you read through my list, please keep in mind that I tried to take notes (well, Granola did most of the job), but I may have missed some details. Enjoy the curated list, and let me know what you think.

Wed 18 Feb

👉 Professional Development Session for New and Aspiring Educators

The first day was packed with wisdom for new and aspiring CS professors (like me). Thanks to Mohammad Azhar and Debasis Bhattacharya for organizing it and bringing all these great CS educators together.

Career Priorities

The opening career panel featured Timothy Yuen, Deepti Joshi, and Chandrika Satyavolu. Each panelist offered a slightly different perspective shaped by their institutional context, but a few themes ran consistently through all three: protect your calendar, don’t overcommit in your first years, and find mentors who will advocate for you (not just advise you).

Timothy emphasized blocking time intentionally for research, teaching prep, and proposal writing, and treated that calendar like a professional commitment.

Deepti, who works in a teaching-focused environment, framed everything around schedule management: teaching, service, and research (nothing else!). She also made a point that resonated with the room: intro to programming has to evolve, even if it’s uncomfortable, because students need to connect with what the industry actually looks like.

Chandrika was perhaps the most candid about the adjustment period, noting that new educators often work evenings and weekends out of necessity, and that having a mentor who understands that reality is essential, something worth asking about directly in job interviews.

Some ABCs of Teaching

Colleen Lewis delivered one of the most practical and energetic presentations of the day. Drawing from her “ABCs of Teaching” framework, she covered a lot of ground quickly and made it stick.

On ability, she urged educators to start from the assumption that all students can succeed, and that learning is genuinely hard and slow, forgetting is normal, and treating it otherwise sets students up to feel like failures.

On body language, the advice was “simple”: smile, move around, and point at things rather than vague directions.

On channels, she warned against slides that compete with spoken explanation: cognitive overload is real, and students can’t listen and read at the same time.

On assessment, she noted that retrieval practice is one of the most effective learning strategies, that students consistently overestimate their understanding, and that frequent low-stakes assessments beat infrequent high-stakes ones every time.

She also addressed boundaries with pragmatism: everything can always be better, so set a stopping point and honor it.

On clarity, she pushed educators to be explicit about learning goals, study strategies, and even whether students should be taking notes during lecture.

On active learning, she made a simple but important point: the buzzword means little without specifics. Don’t just lecture: build in breaks and incorporate group work. The harder part is managing the social awkwardness that group work creates, which she addressed separately with concrete techniques like counting down to bring the class back together and giving students explicit prompts to find and engage their partners.

On anxiety, she suggested identifying your fears and planning around them and offered a reframe worth sitting with: “What if I decided to not be anxious about this?”

On flexibility, she recommended creating rules so you don’t have to make individual decisions in the moment. Offer late days, use a simple form to grant 24-hour extensions so you can track students.

On absorb awkwardness, she offered concrete techniques for the moments that trip up even experienced instructors. Help students find groups by asking them to physically point at their partner. Break the silence with countdowns: “In 10 seconds, everyone will shout their partner’s name... 10, 9, 8...” And warn people before bringing the class back together: “We’ll come back together in 5, 4, 3, 2, 1.”

And on feedback, she offered a suggestion: don’t read your own teaching evaluations: have a TA, or even ChatGPT summarize them, so you get the signal without the emotional noise. She also recommended asking for feedback at the five-minute level and taking notes during class when something doesn’t go well.

She also encouraged educators to avoid reinventing the wheel: email colleagues directly to ask for their slides or assignments, and take advantage of existing repositories such as nifty.stanford.edu, engage-csedu.org (thanks to Brian O’Neill from Quinnipiac University for introducing us to this platform, which provides faculty-contributed, peer-reviewed course materials for a variety of CS courses), as well as peerinstruction4cs.org. She also highlighted csteachingtips.org and the book The ABCs of How We Learn by Schwartz, Tsang, and Blair as foundational reading for any new educator.

Navigating the AI Coding Revolution

Leo Porter‘s session was forward-looking and grounded in real classroom data. He co-runs the GenAI and CS Education Consortium with colleagues at UCSD and the University of Toronto, and has been systematically thinking through what AI means for how we teach and assess programming education.

The industry has been asking for AI-adjacent skills (specification, verification, debugging, navigating complex systems) for decades. Students are already being asked about AI tool usage in job interviews, which means educators who ignore the shift are widening the preparation gap.

His three-pillar framework for adapting was practical. First, assessment reform: more frequent, lower-stakes exercises, clearer per-course AI policies (intro courses may allow it; assembly language courses probably won’t), and a shift toward proctored written or oral exams and verbal check-ins for projects. Second, AI-human tutor integration: AI tools like CodeHelp and CodeAid can handle lower-level support, freeing human TAs to focus on mentoring, career conversations, and higher-order guidance, roles that remain distinctly human. Third, AI literacy as curriculum: teaching students to use AI tools effectively, especially in non-major courses, is itself a learning goal worth designing around.

Leo noted that some courses (systems, architecture, theory) will change very little. Others, especially intro sequences and software engineering, are already in flux. His advice: collaborate with existing course instructors rather than redesigning solo, and don’t drop courses without careful deliberation. You can find more of his work at his Google Scholar page, and the consortium’s resources at teachcswithai.org.

Job Searching and Interviewing

The afternoon job search panel — featuring Sophia Krause-Levy, Timothy Yuen, and Chandrika Satyavolu — tackled the practical mechanics of landing a faculty position. A few pieces of advice stood out:

On application materials: be intentional about the shift from collaborative “we” language to individual “I” — reviewers need to understand your specific contribution, and it’s easy to undersell yourself by habit. Customize your cover letter with institution-specific content (programs, courses, traditions), but keep your core documents stable.

On interviews: teaching demonstrations require real preparation. Request syllabi and student background information in advance, bring backup plans for different knowledge levels, and bring your own watch, since many interview rooms don’t have clocks.

On salary: always negotiate, and do it with specificity. You don’t need another offer to negotiate, and being reasonable and well-researched goes a long way. Build a detailed spreadsheet with itemized requests, cite the Taulbee Report for CS salary benchmarks, and think beyond base salary to course releases, startup funds, moving expenses, and childcare access.

Promotion and Tenure

The tenure panel brought together Leen-Kiat Soh and Christine Alvarado to walk through what the promotion process actually looks like from the inside.

The standard three pillars (research, teaching, service) are weighted differently by institution type, with R1s prioritizing research and teaching-track positions flipping that order. At UC San Diego, the system involves step-based advancement every two to three years, with continuous review even after tenure. The process spans most of an academic year: October submission to May decision, moving through department committee, chair, college review, and university-level approval.

Key advice: start building your dossier three years before your application date, make your research statement about impact rather than technical detail, and keep it readable for non-specialists. Department chair support is essential! And Mohammad Azhar offered a reminder worth holding onto: community colleges are genuinely great options if you love a city and want to invest deeply in your teaching.

👉 Small-School Mindsets at Any Institution — Zach Dodds, Julie Medero, Lauren Bricker

The closing session featured Lauren Bricker (University of Washington), Julie Medero, and Zach Dodds (both Harvey Mudd) sharing approaches to inclusive CS education that can scale beyond small liberal arts settings.

Lauren described UW’s Pathways for Inclusive Excellence program, which combines holistic admissions, a four-week summer bridge, and supplemental workshop support — yielding 85–88% completion rates compared to a 25% baseline for first-generation and low-income students. The model relies on active learning, peer cohort building, bi-weekly small group presentations, and weekly one-on-one TA sessions.

Julie and Zach brought the Harvey Mudd perspective. Julie transformed her mid-level CS course through standards-based grading, self-assessments that drove class focus toward students’ weakest areas, and replacing high-stakes problem sets with optional “challenge problems.” Zach scaled a similar philosophy to 200 students using 15-minute weekly demo sessions with undergraduate TAs, a Google Sheets signup system for 240 time slots, and assignments designed to connect CS to students’ actual lives — a Pokémon API project being a standout example.

Zach is, as noted in the room, a genuinely enthusiastic presence — the kind of educator who makes you want to rethink your own lecture format before you’ve even left the building. If you want inspiration for making CS lectures more engaging, his approach is worth exploring further.

👉 Student Participation in Team-based Software Projects

I only caught the last ten minutes of this session, but it was enough to jot down two resources I’ve since looked through and think are worth sharing. POSSE (Professors’ Open Source Software Experience) is a professional development program for educators looking to integrate open source projects into their courses. They also have a broader learning resources page at teachingopensource.org/Learning_Resources. Limited context on my end, but the links speak for themselves.

👉 Teaching with AI

Later in the day, I stopped by part of the CS50 AI Workshop, led by the Harvard CS50 team. The session opened with an overview of how CS50 has integrated genAI into their teaching since Spring 2023, starting with a course-specific ChatGPT variant and evolving into a suite of custom tools designed around a simple but ambitious goal: approximating a 1:1 teacher-to-student ratio at scale. Rather than allowing generic AI tools like ChatGPT or Copilot, CS50 built their own, including CS50 Duck, a rubber duck debugger that gives instructional responses without handing students the answer. The numbers: 21,000 prompts per day, over 18 million questions answered, at roughly $1 per student per month. I only stayed for the first part (the intro, environment setup, and overview of CS50.ai) but the hands-on portion covering the OpenAI Responses API is available in full on YouTube. If you’re curious about building your own pedagogical AI tools, it’s well worth watching.

Thu 19 Feb

👉 Computer Science Curricula CS2023 Revisited in Light of Generative AI

This panel, featuring Amruth N. Kumar (Ramapo College of New Jersey), Richard Blumenthal (Regis University), Pankaj Jalote (IIIT-Delhi) and Titus Winters (Adobe), took a hard look at whether the CS2023 curriculum guidelines still hold up in light of generative AI and where they may need rethinking.

The starting point was that the traditional CS curriculum has been heavily weighted toward writing code, with comparatively little time spent on reading, testing, debugging, or writing specifications. That balance made sense when writing code was the bottleneck. It is less obvious that it still does. The industry perspective reinforced this shift: the bottleneck has moved from implementation to design. AI can produce a decent implementation in seconds for any well-defined problem, but what it cannot do is decompose a vague problem, write a precise specification, or distinguish between code that merely works and code that is truly good. Instilling that kind of good taste and critical judgment in students was one of the most memorable framings of the session.

On the pedagogical side, the panel discussed experiments with restructuring introductory courses around this tension: placing more assessment emphasis on understanding and diagnosing code rather than simply completing functions; introducing testing early and requiring it throughout; and splitting coursework between proctored work with no internet access and take-home assignments in which students could explicitly use GenAI. The goal is not to ban AI, but to ensure that students develop the skills it cannot substitute.

The ethical dimension came up as well: technology changes; values do not. The ACM Code of Ethics, the panel argued, should be woven into every assignment rather than siloed into a standalone course. Students also need a working understanding of what GenAI actually is.

👉 Teaching with Generative AI: Tools You Can Use Today

This session organized by the ACM Task Force on Generative AI and Programming Assessment and introduced by Paul Denny and Juho Leinonen opened with survey findings from 700 educators globally: 74% want professional development for AI integration, and 87% worry about student over-reliance.

Six tools followed:

Mark Liffiton presented CodeHelp, an open-source AI tutoring tool with guardrails that prevent direct code answers. Instructors design the tutoring plans and learning objectives, and the drag-and-drop interface makes setup surprisingly straightforward.

Kristin Stephens-Martinez shared a structured brainstorming activity using LLMs for group data science projects. Students work with a provided prompt template and reflect on the process afterward. The goal is authorized, accountable AI use rather than avoidance. Her slides and a blog post are available if you want to dig in.

Rahad Arman Nabid presented Autosuggestion Quiz, which addresses over-reliance on Copilot-style tools by putting students in a simulated IDE where they select from multiple code suggestions (correct and wrong) one line at a time. The distractor injection process is particularly clever, and it logs student interaction data throughout.

Sverrir Thorgeirsson demoed Ente, an online IDE tested with over 1,000 students at ETH, with customizable AI agents, cost controls, and detailed analytics.

Anastasiia Birillo presented ANVIL, a tool from JetBrains and TU Delft that generates short pedagogical videos explaining CS concepts, with a human-in-the-loop verification step before students see anything.

David H. Smith closed with Purplex, a platform built around explaining code in plain language: students describe what code does in natural language, and an automated system grades the quality and abstraction level of their descriptions. It also works multilingually.

Fri 20 Feb

The third day opened with the Awards Ceremony, presented by Claudia Szabo, followed by an awards panel on the future of CS education moderated by Mark Sherriff.

Three awards were given. Richard Ladner from the University of Washington received the Outstanding Contribution to CS Education award for decades of work developing technologies and activities focused on computational literacy for discovery and expression. The Distinguished Service Award went to Paul Tymann, recognized for thirty years of dedicated advocacy for students and faculty across a diverse community. And the Test of Time Award went to Philip Guo for “Online Python Tutor,” an embeddable web-based program visualization tool published thirteen years ago that remains widely used today.

The awards panel that followed turned into a broader reflection on how computing education has evolved and what remains unresolved. The microcomputer revolution of the late 70s and early 80s came up as a major inflection point: the moment computing moved out of the mainframe and into people’s hands. On the current landscape, the conversation touched on AI being better treated as a tool than a solution. There was also an honest acknowledgment that many CS educators never attend SIGCSE and don’t know this community exists. Both honorees defined success similarly: not by publications or grants, but by the students whose careers they had shaped along the way.

👉 Career Paths in Tech: What Students Do, Post, and Code to Get Hired

This session on CS internships and career pathways was chaired by Gagan Garg, who kept the energy up throughout. One of those chairs who makes a session noticeably more enjoyable just by being in the room.

Four papers were presented:

Christopher Perdriau (currently on the job market) opened with a large-scale study of over 14,000 undergraduate CS majors across 408 institutions, looking at what predicts internship attainment. Hackathon participation had the highest impact, increasing internship probability by 11%, followed by leadership roles in CS groups and attendance at conferences. Students with disabilities, first-generation students, and non-US citizens were significantly less likely to secure internships despite participating in extracurriculars at comparable rates.

The second paper examined TikTok as an informal career guidance resource for CS students. Analyzing 200 videos across the top career-related hashtags, the study found heavy skew toward software engineering roles, with half of videos focused on salary and company prestige. Actionable technical advice was scarce, and the gap between what the platform shows and what the job market actually looks like raises real questions about how students are forming their expectations.

Utsab Saha presented work from the Computing Talent Initiative and CodeDay, a program that has helped over 4,000 students make their first open source contribution, backed by 1,400+ industry mentors. The most interesting part for me was a structured process developed with hiring managers: define your hiring bar, map it to demonstrable open-source activities, and sign agreements for student referrals. The top qualities those managers identified were problem-solving, initiative, and curiosity (domain knowledge came last!). Tyler Menezes also deserves a mention for his contributions to making this work possible.

Stacey Levine closed with CS Pathways, an interactive tool that maps courses to careers and careers to elective recommendations. The problem it addresses is that students often default to software engineering because it’s what they know. In a study of 25 graduate students, 100% discovered careers they hadn’t previously considered, and 36% said they were considering changing direction.

Sat 21 Feb

👉 CS and SE Education, post-AI

Titus Winters, Senior Principal Scientist at Adobe and author of Software Engineering at Google (the “Flamingo Book”), closed the conference with one of my favorite talks of the week.

He began by noting that CS unemployment rates now mirror those of performing arts graduates. Tech hiring has stagnated, and AI coding tools are delivering real productivity gains in industry, around 10% by Google’s own numbers. The question he posed to the room was whether CS education is preparing students for that reality, and his answer was largely no.

He started with the curriculum. He argued that CS programs spend too much time on implementation details that no longer reflect how modern computing works. Binary search trees get a week of lectures despite hash tables being superior in most real-world scenarios; recursion is taught without addressing tail call optimization; regular expressions are covered without addressing their exponential edge cases. The deeper problem, borrowing a 1983 critique by Mills, is that we teach the “French dictionary” instead of “French poetry,” meaning isolated concepts rather than holistic program quality.

Modern computing, he argued, has largely converged on two container types for the vast majority of use cases: contiguous sequence containers and associative containers. Memory access patterns now matter more than algorithmic complexity, and average-case performance plus cache effects often beat worst-case Big-O analysis in practice.

His proposed direction is to start CS majors with code understanding before code writing, grade the process and not just the output, including code, prompts, tests, and peer critique, and treat assignments more like art projects where quality and elegance matter. For non-majors, he proposed a CS0 course focused on AI-assisted coding with safety awareness. For the field more broadly, he said the focus should be on what humans still do better than AI: ethics, communication, architecture, and the kind of taste and critical judgment that come from real critique and practice.

He closed with a direct challenge to junior faculty saying academic CS risks losing relevance to bootcamps and software engineering departments if it doesn’t adapt but that same moment is an opportunity to lead other disciplines through the AI disruption rather than be left behind by it. The path forward is real, but it requires letting go of legacy concepts that no longer serve students.

👉 Codeless Computing: Boolean Logic, Parallelism, and ML for Actual Humans

I arrived late to this session and missed Sverrir Thorgeirsson’s presentation on a code-free interface for constructing Boolean expressions, but his paper is worth reading if the topic interests you.

Chris Bourke from the University of Nebraska–Lincoln presented NSF-supported work on bringing parallel and distributed computing into the early CS curriculum. The problem is that the last major advancement in single-core processors occurred over thirteen years ago, yet parallel computing remains largely absent from CS1. His approach uses code-free, web-based visualizations, with no programming required, to introduce concepts such as synchronous versus asynchronous execution, embarrassingly parallel applications, and producer–consumer patterns. A study across four CS1 courses showed statistically significant improvements on all eight assessment questions in the intervention group, with medium to large effect sizes, while the control group showed minimal gains.

Yaniv Yacoby from Wellesley College closed with a course on probabilistic machine learning designed for liberal arts students. It has low prerequisites, uses a probabilistic programming language to avoid complex derivations, and presents a unified framework that connects methods such as linear regression and K-means through directed graphical models. The course is structured around three layers of AI bias: methods and evaluation metrics, data quality and collection, and the social and moral critique that is often left out. Students reported feeling more equipped to participate in AI conversations and better able to question assumptions. The full 24-chapter open textbook is available at mogu-lab.github.io/probabilistic-foundations-of-ml.

Thanks to Mohsen Dorodchi from UNC Charlotte for chairing.

Before closing, a few more general thoughts on the conference:

The Computing Education community is simply one of the best communities I’ve been part of. Friendly, open, with no egos and no gatekeeping. The kind of place where you end up in a genuinely good conversation with a stranger between sessions and walk away with a new collaborator or just a good memory. I met wonderful people including Thomas Rexin, Isabela Figueira, Kush Patel, Kelly Lindsay, Ksenia Shneyveys, Aizen Baidya, Anushka Joshi, Tommaso Carraro, Sheila Foley, Amrita Ganguly, Carter Ithier, Jacob Roberts-Baca, Pedro Guillermo Feijóo-García, Rifat Sabbir Mansur, Jhonathan Sora Cárdenas, Houyame Lkhider, and many more.

Volunteering was great, and I’d recommend it to any PhD student. There are quiet stretches where you end up talking with other PhD students, faculty, and junior CS students in ways that wouldn’t happen otherwise. The hours are manageable, the mentoring moments are genuine, and it’s worth every bit of it.

Some of the best conversations happened in the margins — with Juho, Anastasiia, and others. Excited about what comes next from those exchanges.

Two downsides: no Wi-Fi and no catering on Wednesday. The Harvard CS50 workshop was great but ran past 10 pm, which was a lot to ask of people who had early starts the next morning. I understand the logic of evening slots, but by that point, the energy just wasn’t there to process it well.

If you’re interested in reviewing for SIGCSE TS 2027 — which will be in Sacramento, CA — the form is open.

It’s been a wonderful few days. I’ve had great moments and met amazing people. Thanks to everyone who made it what it was and a special thanks to my advisor Amin for making the trip possible.

🔍 Resources for Learning CS

No time this week to look for additional learning resources beyond the ones shared in the SIGCSE TS reflection. I’ll look for more next week.

🔍 Resources for Teaching CS

No time this week to look for additional learning resources beyond the ones shared in the SIGCSE TS reflection. I’ll look for more next week.

🦄 Quick bytes from the industry

→ David Ronca’s Career Journey (Netflix + Meta)

After decades leading video engineering at Netflix and Facebook/Meta, David Ronca now speaks with the clarity of someone who has seen both theory and messy production systems up close. In my view, his most compelling reflections focus on the gap between academic excellence and real-world engineering, and on the crucial difference between interviewing well and actually building well.

1) Engineering over academic purity: One of the most interesting threads in Ronca’s story is his relationship with academia. He deeply respects academic rigor—some of the strongest engineers he worked with had academic backgrounds—but he consistently frames real-world engineering as something different from theoretical elegance. His long collaboration with Ioannis Katsavounidis is a good example: Ioannis came from a signal processing and academic background, but his real impact came from forcing academic models into production reality at Netflix and Meta scale. It wasn’t enough for a codec to look good on two videos in a paper; it had to hold across millions or billions of videos under real constraints. Ronca’s view is not anti-academic—it’s anti-theory-without-practice. The real test of brilliance is whether it survives production, scale, and messy constraints.

2) Interviewing well vs. building well: Ronca is openly skeptical of how much conventional technical interviews—especially LeetCode-style screens—actually tell us about engineering ability. In his experience, many candidates master these interviews, solve algorithmic problems quickly, and pass structured system design rounds, only to struggle later with real engineering decisions. LeetCode, in his view, measures preparation and speed under artificial constraints; it does not reliably measure critical judgment. What he cares about instead is whether someone can reason about complex systems, understand trade-offs, and make sound decisions with incomplete information. Those qualities are harder to standardize and harder to test at scale, but they are far more predictive of long-term impact. The danger, he suggests, is confusing interview performance with engineering depth—and optimizing hiring processes for what is easy to measure rather than what truly matters.

→ Ben Thompson on AI ads, the end of SaaS, and the future of media

I was listening to Ben Thompson on Cheeky Pint during my flight to St. Louis. Great stuff.

🌎 Computing Education Community

ACM members are invited to nominate themselves or colleagues (in good standing) as At-Large Members of the Education Advisory Committee (EAC) by March 26, 2026 (AoE) via this link, with selected nominees invited to attend the August 2026 EAC meeting (in person or remotely).
The University of Arizona Department of Computer Science is hiring.
The UK and Ireland Computing Education Research (UKICER) Conference will take place in Cambridge on September 3–4; paper abstracts are due April 13 and full papers April 21 (see the website for other submission deadlines).
University of Houston (Bauer College) is hiring a Professor of Practice in Artificial Intelligence (non-tenure track) starting Fall 2026, seeking candidates with a PhD (or equivalent) and strong AI/ML/data science expertise—ideally with industry experience—to teach and help shape AI curriculum in business-focused contexts; review begins April 1, 2026.

🤔 Thought(s) For You to Ponder…

Each AI tool also creates new demands: more tasks, more moving parts, blurrier roles, and greater burnout. While it’s true that it promises to free us from certain responsibilities, it ultimately ends up creating new ways to fill our time. This HBR article shares a practical case study with additional data to help us think more critically about how to integrate AI thoughtfully, temper expectations, ease up on the gas, and set clear boundaries so we don’t end up burning ourselves out. Related to this, Armin Ronacher argues that the new bottleneck is our human ability to truly understand—and take responsibility for—the code generated by AI, which is now being produced at a much faster pace.

Lovely. No explicit lessons. No agenda beyond sharing a love for computing and showing how mentorship, experimentation, and the passage of time shape a 25-year career devoted to this craft. Thank you for sharing this, Susam Pal. Truly inspiring.

If he’s not on your radar yet, Andreas (best known as the former CTO of Product Hunt and later VP of Engineering at CoinList) is a standout in the startup world and one of those people who’s just genuinely great to listen to. More recently, he’s been investing through his solo GP fund, PROTOTYPE. Here’s his YT channel.

AI news: the creator of OpenClaw is joining the OpenAI team. A few months ago, he was all about Claude Code. In fact, his project originally launched under the name Clawcode, but Anthropic sent him legal letters over the name and spooked him. Peter then started using Codex and migrated his infrastructure over there. Sam Altman clearly saw the opportunity. Let’s see how this plays out.

There are some very interesting things in here: The future of software development retreat. AI is pushing software engineering toward guiding agents, writing clearer specs and tests, and thinking more about risk, while raising concerns about security, growing system complexity, and blurred roles.

There’s always something valuable to learn from anyone — even from those who criticize us or see things differently.

Sonnet 4.6 is here.

Very good interview: Dario Amodei of Anthropic on the Dwarkesh podcast

Quote: What has been the most surprising thing is the lack of public recognition of how close we are to the end of the exponential. To me, it is absolutely wild that you have people — within the bubble and outside the bubble — talking about the same tired, old hot-button political issues, when we are near the end of the exponential.

I’d really like to read a technical article about Spotify’s internal system, called Honk, which is built on top of Claude Code. Co-CEO Gustav Söderström credits this system with helping the company launch more than 50 new features in 2025. He also said that their “best developers” (who exactly would that be?) haven’t written a single line of code since December. Pretty intriguing.

Why Physical Making Matters (Austin Kleon): We are creatures, and it’s cool to lean into our creatureliness and feel like full human beings.

📌 Research Corner

I’ve been doing a full-on literature review all week. It’s exhausting to go through so many papers, but it really helps clarify things from different angles: matching conferences to keywords, identifying the most active researchers in the field, finding the most cited papers, and discovering interesting academics to reference. My new project focuses on Parsons Problems and misconceptions in OOP. I’m also interested in behavioral traces and scaffolding. I’m open to suggestions or examples!
It applies to research as well. Good judgment is shaped through critique. I also believe that collective feedback and thoughtful input from your research lab peers isn’t about reaching consensus. It’s about uncovering blind spots, bridging skill gaps among teammates, and sharpening the judgment of the researcher presenting the work. The responsibility still lies with the presenter, but the project improves by being tested in public.

🪁 Leisure Line

Gotta love Shipley’s Donuts! The chocolate-iced donut with walnuts is delicious.

📖📺🍿 Currently Reading, Watching, Listening

This animation by Ousmane Cissé is truly beautiful. It’s a tribute to the characters from The Legend of Zelda, the Nintendo game. You can check out the piece here:

This week I finished the audiobook Sobre Dios (On God) by Byung-Chul Han. It’s really thought-provoking—very much in his style—and it gives you a lot to chew on and dive into regarding the structural aspects of how we live that make experiencing the divine so difficult. I like the way he raises questions and how he articulates them. You’ll enjoy it if you’re into contemporary philosophy and are interested in a non-confessional take on religious experience.

🌐 Cool things from around the internet

🔗 Nova — an infinite canvas for your personal development.

🔗 Ferrari Luce — first electric Ferrari.

Issue #33 of Computing Education Things was written while listening to:

🔗 Quick Links

🎧 Listen to Computing Education Things Podcast

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

Buy Me A Coffee

#32 — AI can’t add flavor to Computing Education

Daniel Prol — Fri, 13 Feb 2026 11:20:19 GMT

Reflections

Salt and AI

The Salt Cathedral is an impressive underground Catholic cathedral located in Zipaquirá, Colombia. Credit.

In Matthew 5:13–16, Jesus encourages his disciples to be salt in the midst of the world. I’ve always found that image powerful—salt is what gives food its flavor. In family life, in our relationships, in the classroom, in our daily work—in all things—we each have the ability to add that kind of flavor to the moments we share. And we do it in our own way: through our personality, our sense of humor, our presence, and our human values.

In today’s AI-centric world, we may need those “human ingredients” more than ever: joy, humor, deep listening, humility, collaboration, empathy, attention to small details, a word of comfort, the right advice at the right time. These are things AI simply cannot offer. We can sense that AI, for all its power, lacks flavor. Only we can bring depth and richness to Computing Education. Think of the local store customers choose not because they can’t buy the same products online, but because of the warmth of the service, the care, the personal attention—the sense that everything is done with love.

This “salt” is what restores flavor for those who feel disengaged by AI, helping preserve integrity not just in how we think, but also in how we act. Often these values go unnoticed, like a pinch of salt you never see but that transforms the whole dish. The same is true in Computing Education. Our pedagogy should cultivate the human values our moment so desperately needs. But to cultivate them in our students, we must first live them ourselves. If our classes have more flavor because they are engaging, because we serve with heart, carry a good attitude, and help each student leave with greater clarity than they came with—that alone makes the work worthwhile.

Only a few days remain before the deadline to apply for WG10 at ITiCSE 2026 (Teamwork in Computing Education: Skills, Values, and Virtues)— I’ve already applied!

A Lesser-Known Bottleneck in AI Scalability

In this week’s Numerical Analysis class, we covered interpolation and fitting, and it turned into a good opportunity to think about a question that keeps coming up in AI conversations: scalability.

That’s where this article by Paula Lamo, a former professor of mine from a Computer Technology course, fits in nicely. She explores a lesser-known bottleneck that’s increasingly shaping the limits of AI. It’s not just about how much electricity is consumed, but about a more subtle and deeply technical constraint: power electronics. As power density rises in the accelerators used to train and run AI models, managing and converting energy becomes a critical design challenge. Inefficiencies at this level lead to heat, instability, and hard physical limits that software alone can’t overcome.

Many of the infrastructure and cost issues we see today trace back to this very specific physical constraint. I found this piece especially valuable because it pushes back against the dominant narrative that places most of the emphasis on data and code, and instead highlights how the future of AI is increasingly constrained by physical and engineering realities.

Those who have lived in software land don’t realize they’re about to have a hard lesson in hardware - Elon Musk on the Dwarkesh podcast.

What skills do CS students need?

In the new job market shaped by AI, technical fluency alone is no longer enough. We need to think of AI not just as a tool, but as an “e-bike for the mind”—something that can extend our range, but also magnify our mistakes. In my opinion, it is important for CS students to develop four key skills:

Technical fluency: knowing how to use AI tools effectively, but also understanding how they work (and how they fail).
Ethical reasoning: the ability to wrestle with real trade-offs and to think seriously about societal, environmental, and human consequences.
Creative agency: using AI to expand human capability rather than outsourcing thinking to it—maintaining ownership over ideas, design, and direction.
Critical judgment: knowing when to trust, verify, or reject AI outputs, and even when not to build something at all.

If we want to prepare the next generation of computer scientists thoughtfully, CS educators should make space to intentionally teach and model these four capacities—not just as complementary topics, but as important parts of a well-rounded formation.

🔍 Resources for Learning CS

→ Rendering ASCII art

A technical article on high-fidelity ASCII rendering by Alex Harri that’s also a beautiful exercise in craft, full of 6D vectors, directional contrast tweaks, and a satisfying obsession with crisp edges.

→ Building a Trustworthy AI Data Agent

This post explains how OpenAI built an internal AI “data agent” that lets employees analyze large datasets using natural language. It’s a concrete look at what it takes to make AI genuinely useful for day-to-day analytics while preserving reliability, debuggability, and trust.

→ Find your next chart inspiration using AI

This project lets you search TidyTuesday contributions, Datawrapper’s Data Viz Dispatch, and FlowingData using natural language to find examples of specific visualization types, encodings, chart elements, statistical methods, datasets, and more.

→ How is data stored?

Dan Hollick explores how data is physically stored in computers, explaining the trade-offs between technologies like SRAM, DRAM, SSDs, and HDDs. He shows how memory architecture balances speed, capacity, power, and permanence—and why software is ultimately constrained by hardware design decisions.

🔍 Resources for Teaching CS

→ Accessibility for everyone

Designing for accessibility is a valuable skill in UX, and if you teach HCI and cover accessibility, you can now read Accessibility for Everyone by Laura Kalbag for free online.

→ Thinking critically about AI literacy

I loved this episode on thinking critically when someone is talking about AI literacy. It points out historical parallels with web literacy, where we didn’t know what it was and taught students very bad practices instead of waiting for research to determine what it should be.

→ Open Visualization Academy

I love this project. I think Alberto Cairo is a clear example of how an academic can go beyond the classroom and make a real impact by sharing knowledge through online academies like this one, while also building a strong sense of community among fellow academics and data visualization professionals.

🦄 Quick bytes from the industry

→ Adam Ernst’s Career Journey (Meta IC9)

The Developing Dev

Meta Distinguished Eng (IC9) On Influencing Engs, Failures, and Learnings

This is Adam Ernst, a Distinguished Engineer at Meta (IC9) who’s built iOS infrastructure that has impacted the entire company. We talked about how his career grew, a major failed project of his, and everything he learned growing to that level…

Listen now

4 months ago · 21 likes · 1 comment · Ryan Peterman

Adam Ernst graduated in Computer Science from Princeton in 2010 and joined Meta in 2012, just ahead of the company’s IPO, during the mobile rewrite era. He has since spent more than 13 years at Meta, where he is now a Distinguished Engineer (IC9), building and leading critical iOS infrastructure with company-wide impact. In this conversation, Adam walks through his career—from early projects to ambitious platform bets—sharing concrete lessons on growing as an individual contributor, influencing without authority, and learning from both success and failure.

Early Builder Mindset

Adam’s trajectory starts long before Meta: as a middle-schooler he built and sold a real product (“Cosmic Soft”)—online testing software for teachers—using a beginner-friendly tool (RealBasic). He even handled payments through a proto-Stripe service (eSellerate) and accepted checks by mail, emailing license codes back. My takeaway: build something useful, get it into the world, learn by shipping.

Scaling iOS at Facebook

He joined Facebook in 2012 (right before the IPO) during the shift from an HTML5 app to a native rewrite—an environment where scaling problems created unusually large opportunities. His early impact came from replacing Apple’s CoreData, which didn’t scale to a rapidly growing iOS org. The key lesson: architect for scale while enabling incremental adoption (e.g., immutable “memmodels” to reason about thread safety and change safely across a fast-moving codebase).

Influence without authority

A recurring theme is how Adam persuades other software engineers without formal power: talk live when possible, acknowledge the other side’s worldview (“I like vanilla Apple frameworks too”), and arrive with data (even reverse-engineering black-box behavior when needed). His most effective tactic: do the work for people—show up with a migration or fix already implemented so the task becomes “approve” rather than “take on a huge task.” He applies that same influence model through code review, treating it as a concrete forum for technical dialogue—using real diffs to explain trade-offs, surface assumptions, and shape how other software engineers reason about code.

High-stakes platform bets

Adam’s senior-career arc is shaped by high-risk, platform-level bets. With ComponentKit, he helped introduce React-inspired declarative UI ideas to iOS years before SwiftUI or React Native, addressing News Feed complexity through components, immutability, and better performance—while navigating significant internal skepticism and the need for sustained buy-in across teams. The counterweight was ComponentScript, a cross-platform framework that was technically sound but failed to gain traction. Adam is explicit about why: unclear target users, real adoption friction, and ecosystem dynamics mattered as much as the quality of the architecture itself. His takeaway is clear and hard-earned: strong technical ideas still fail without alignment, incentives, and momentum—and senior engineers must be willing to shut projects down responsibly when they don’t work.

Technical Depth

Rather than chasing trends or frequently switching domains, Adam deliberately stayed deep in mobile infrastructure, allowing real problems to pull him into adjacent systems when needed. When blocked by GraphQL, build tooling, or codegen, he didn’t escalate—he dove in, traced the issue several layers deep, and either fixed it himself or arrived with a precise diagnosis. Over time, this pattern compounded into broad system knowledge rooted in necessity. Combined with his bias toward writing and reviewing real code, that depth became his leverage as an IC: influence earned through proximity to the work, not through titles or constant reinvention.

→ Bots, Agents, and Leverage

In this conversation with Aman Manazir, Ritesh Verma explains how he went from a traditional computer science path (University of Maryland, internship and full-time offer at Capital One) to making more from side hustles than from his six-figure software engineering job. His journey started with experimentation: a viral SAT YouTube channel in college, then niche automation projects like checkout bots that attracted high-paying clients.

Over time, he realized that what people now call “AI agents” are often just automation software with an AI layer—the evolution of the same bots he had been building for years. To him, the technical barrier is lower than people think. Tools like no-code automation platforms, web automation libraries, and AI-assisted coding environments mean that most software engineers already have more than enough skill to compete.

Computer science students are the best people to do this because of their unique advantage over anyone else.

The real differentiator is identifying a clear problem, packaging a solution quickly, and getting it in front of buyers. He repeatedly emphasizes that many profitable agents are simple: AI chatbots trained on a company’s website to capture leads instantly, automated content systems for e-commerce brands, or niche workflows that save small businesses hours of manual work. His playbook is deliberately practical: start with a niche you can access, find a time-consuming process, build a live demo, price it low enough to remove friction for the first customer, capture a video testimonial, and turn that into a landing page. He stresses that credibility compounds: once you have one paying client and proof of results, the second and third come much faster. Instead of building a perfect product in isolation, he advocates building in public and using real-world case studies as marketing fuel.

Distribution, in his view, is the true moat. Short-form content can generate attention quickly. X (Twitter) allows builders to tap into “build in public” communities. But he highlights Reddit as especially powerful: transparent posts about revenue milestones or lessons learned—without links, calls-to-action, or overt selling—can drive serious inbound leads. If the story is compelling enough, prospects will find a way to reach out.

Underneath the tactics is a broader philosophy. Ritesh sees this path not as hype, but as economic self-defense in a market defined by layoffs, shrinking junior roles, and increasing automation inside large companies. In his view, software engineers should treat side hustles not as hobbies but as leverage—cash-flowing systems that reduce dependence on a single employer. And if someone chooses to leave a salaried role, he suggests a simple rule: don’t quit for the dream—quit when the business reliably earns two to three times your salary.

🌎 Computing Education Community

Khoury College of Computer Sciences at Northeastern University is recruiting a Director of Computing Programs (Associate/Full Teaching Professor) for its Northeastern University Oakland Campus, starting July or September 2026 (apply via HigherEdJobs).
A new community on Epistemic Programming—programming for gaining insights—is forming! ITiCSE 2026 Working Group 7 is now recruiting participants. Apply by February 16 to join this in-person collaboration, taking place July 10–12 in Madrid.
The Florida International University College of Engineering and Computing is hiring a tenured/tenure-track Associate or Full Professor in the Multidisciplinary Engineering & Computing Education, Systems, and Management Department, with a strong interest in computing education (review begins today Feb. 13).
If you’ll be in St. Louis for SIGCSE Technical Symposium 2026, seats are still available for the affiliated event “Innovations and Opportunities in Liberal Arts Computing Education”.
The University of Pittsburgh Department of Computer Science is hiring two Teaching Assistant Professors to support in-person and online teaching—especially for adult learners.
La Salle University is hiring a Non-Tenure-Track Assistant Professor in Computer Science starting Fall 2026 to teach across CS, IT, Cybersecurity (and possibly AI/CIS graduate programs), with application review beginning Feb. 20, 2026.
ICER 2026 deadline reminder: abstracts for Research Papers are due February 20, 2026.
Koli Calling 2025 highlights: the keynote by R. Benjamin Shapiro, “The Coin Has Three Sides: Human–Computer Symbiosis in the Future of Computing Education,” set the tone, with Best Paper awarded to Henriikka Vartiainen & Matti Tedre, Best Presentation to Naaz Sibia et al., and Best Poster to Radu Mariescu-Istodor & Anssi Gröhn.
The California State University, Fresno Department of Computer Science is hiring a tenure-track Assistant Professor starting AY 2026–27, with priority areas including Computer Science Education, Cloud Computing, Theory, Cybersecurity, and AI (review begins March 1, 2026).
If you’re attending the SIGCSE Technical Symposium 2026 in St. Louis, consider joining the ACM Committee for Computing Education in Community Colleges for the CS Transfer 2Y Curricula Focus Group on Saturday, February 21, from 3:30 to 6:30 PM.
The ACM Education Board invites you to complete a brief survey on first-year CS student preparedness to help inform community-wide recommendations.
CSEd research students are invited to the CSEdRStudent Network online meetup on Feb 19, 2026 (6 pm GMT) to discuss Open Research with Laurie Gale from the Raspberry Pi Computing Education Research Centre (message Katharine Childs or Nicola Looker to join).
This Spring instructors can pilot an NSF-funded web app that brings peer instruction to asynchronous courses: integrating the tool into a class, collaborating with the research team, and receiving a $500 VISA gift card in appreciation for their participation.
Instructors attending SIGCSE are invited to a Feb 18 focus group (4–5 pm CST) on AI-augmented, performance-based assessment. Participants will receive a $100 gift card after the session and brief survey.
Westminster College seeks a tenure-track Computer Science faculty member (AI/ML focus) starting August 2026; full consideration by March 6, 2026.
Join SIGCSE Reads in St. Louis next week—attend the Friday 10:40 a.m. panel in Room 276 on persisting and empowering in CS education, explore this year’s selections (including Whistleblower by Susan Fowler, Service Model by Adrian Tchaikovsky, Annalee Newitz’s short story “When Robot and Crow Saved East Saint Louis,” or the Modern Figures Podcast), and share your ideas for next year’s read via the survey.
Kai Williams and Timothy B. Lee (Understanding AI) are looking to speak today with roboticists and software engineers/product managers about robotics and the recent progress of coding agents. Sign up for a time slot on their calendars (robotics | coding agents).
The Department of Computer Science at UMD is hiring a faculty member to lead curriculum innovation and course development focused on the impact of AI tools in computer science education, alongside a standard teaching load.

🤔 Thought(s) For You to Ponder…

What Alejandro Mora says in afueradentro ties into what I talked about in last week’s newsletter—about drawing from other disciplines so we don’t end up living in a bubble: Having a hobby changes your Instagram algorithm, so you’re not constantly seeing the same stuff. For anyone working in the cultural or creative industries, it’s important to be immersed in different things.

Interesting thread by Boris Cherny with some tips on how to get the most out of Claude Code. Along those lines, I loved this article by Mitchell Hashimoto about his journey adopting AI into his workflow.

Heard here: Andrej Karpathy is a great professor — it’s hard to find someone so technical who can explain things so clearly.

A highlight from ACM’s interview with Pei Cao, VP of Engineering at YouTube: Computer science is a fast-changing field. Every decade brings new challenges that require communities of researchers to solve. As you advance in your career, you won’t always work in the technical subareas in which you were trained. But a solid computer science foundation will enable you to learn any subfield of computer science quickly and then make contributions. So be prepared to learn continuously and don’t be afraid to venture into new areas.

Brendan Gregg is joining OpenAI. What a gig! There are very few places right now where the link between system performance and real business value is as strong as it is there.

I loved this piece from Julian Lehr, Creative Director at Linear, laying out a thoughtful case against conversational interfaces.

julian.digital

The case against conversational interfaces

Intro Conversational interfaces are a bit of a meme. Every couple of years a shiny new AI development emerges and people in tech go “This is it! The next computing paradigm is here! We’ll only use natural language going forward!”. But then nothing actually changes and we continue using computers the way we always have, until the debate resurfaces a few y…

a year ago · 52 likes · 6 comments

In the new version of Mozilla, set to release on February 24, you’ll be able to turn AI tools on or off. It looks like they’re trying to stand out from the competition by giving users the option to have a more controlled experience—based on their own preferences.

Thought-provoking piece from Josh Brake on what Cosmos calls the philosopher-builder.

At Mudd, the humanities, social sciences, and the arts are integrated throughout the curriculum. Students take them throughout their degree alongside their technical courses. Mudd’s core curriculum exposes students to fields across science, engineering, and mathematics. Breadth is a feature, not a bug.
If the first duty of the educator is formational, the question we need to ask is not “what are they learning?” but “who are they becoming?” What habits and practices are we helping our students to cultivate? What questions are we teaching them to ask?

The Absent-Minded Professor

Cultivating Philosopher-Builders

Thank you for being here. Please consider supporting my work by sharing my writing with a friend or taking out a paid subscription…

3 months ago · 10 likes · 2 comments

A great quote from Kevin Kelly that touches on the paradox of choice:

We have not yet and never will, make a technology that we cannot abuse or weaponize. And I’ve been saying this for a while saying, Oh, and by the way, the most powerful technology that we just invented, the internet, we’re going to weaponize and we’re going to abuse it. It’s going to be abused powerfully. And this is the thing, the more powerful the technology, the more powerfully it will be abused. That’s the nature of it. AI, man, it will be really abused. However, and this is the curious thing, even those abuses of technology are increased choices. When the first humanoid picked up a rock and turns it into a hammer, either to make a shelter or to kill his brother, he suddenly had a new choice he never had before. That choice is good.

This piece from Every introduces compound engineering, a model where AI writes the code and software engineering centers on a four-step loop: plan, work, assess, and compound. Its key idea is that value now lies in planning and review, while accumulated agent learning makes each new feature easier to build.

Brilliant video by Jared Henderson tracing how our attention was monetized and arguing we should reclaim it as something we consciously choose to care about.

Looking back to when I wrote about the Orchestrator of skills in this post, this week Eduardo Diaz dives deeper into this new orchestrator role—how to know what to ask for, what context to provide, and when to step in…

VS Code extensions with sidebars are gradually becoming obsolete due to the new agent-based CLI workflow.

Hinton is Second Scientist With Over 1 Million Citations.

📌 Research Corner

This week, I attended two computer science seminars at UH. One of them presented by Gen Li, a third-year Ph.D. student at Clemson University, where he focused on the evolution of modern AI models and the growing challenges related to efficiency, safety, and real-world deployment and where he shared concrete research on dynamic sparse training, selective unlearning, and adversarial defenses—techniques that help make large models more practical, robust, and trustworthy under real-world constraints. The second seminar, given by Jerry Yao-Chieh Hu from Northwestern, took a more theoretical turn, offering a unifying view of what transformers actually do. He framed transformers around two core capabilities—memory and procedure execution—showing how attention can be interpreted as a physical memory retrieval process and how a single fixed model can act as a universal program executor when prompts are treated as programs. The talk connected deep theory with large-scale scientific applications in genomics and astrophysics, shedding light on why foundation models generalize so effectively across tasks.
I learned a lot from these two potential UH tenure-track faculty members.

Next week, from February 18 to 21, I’ll be attending the SIGCSE TS 2026 conference in St. Louis, Missouri. It’ll be my first time attending as a PhD student. If you’re around, feel free to say hi! I would love to talk computing education, AI in programming, etc.

Found a solid example of how clearly articulating the process can be powerful marketing for research.

Gemini is becoming my research partner as well.

What AI tools do you use for your literature review? If you’d like to see mine, I’ve put together a list here:
Google Scholar Labs
alphaXiv
Elicit
Consensus
Scite
searchthearXiv
SciSpace
Semantic Scholar
Connected Papers
arXiv Xplorer
Emergent Mind
Litmaps
Paper Digest
PaperMatch
ResearchRabbit
Paperscape
Asta
Papiers
ChatPDF
AnswerThis
dblp
LLMs: Claude, ChatGPT, Gemini, Manus AI, Perplexity

🪁 Leisure Line

Great weather and blue skies are back in Houston. Nothing beats coffee with a view and a good conversation with Nick Anderson (PhD at UH CS) about life, research, and what’s ahead.

📖📺🍿 Currently Reading, Watching, Listening

This week, I watched Ordinary Angels—a beautiful movie based on a true story. Highly recommended!

Great podcast in Spanish: Chisme Corporativo—it has a really fun storytelling style to share the story of a company. It reminded me of Acquired Podcast. You end up learning the company’s history almost without realizing it. It’s hosted by two Mexican women with that sort of “posh” tone, which I think actually makes the content even more engaging. Plus, both of them have entrepreneurial backgrounds, and it really shows when they speak.

These two men are among my favorite living thinkers, so I absolutely loved their recent conversation.

If, as Brooks puts it, “all the things we care about in life are complex,” then learning computer science must be complex too. You don’t “solve” a marriage once and for all—“I’ll never solve my marriage. I can only live in my marriage.” In the same way, you don’t solve CS and move on; you live in it, wrestle with it, grow through it. The confusion, the bugs, the conceptual friction—those aren’t obstacles to learning. They are THE learning. In complex domains, you cannot skip the difficulty.

I highly recommend the BBC podcast Desert Island Discs, where each guest is asked to imagine they’re stranded on a deserted island. They have to choose eight albums, one book, and one personal item. Through their choices, you get a deep, revealing portrait of the castaway featured in each episode. What we listen to, what we read, and the story behind those choices say a lot about who we are!

🌐 Cool things from around the internet

🔗 Archivio Grafica Italiana — the first online archive dedicated to the entire Italian graphic design heritage.

🔗 25 years of Wikipedia — do you know how Wikipedia started?

🔗 S.C. GJØA — how Athletics rebuilt the brand of New York’s oldest youth soccer club.

🔗 Moody — a smart prompter in your Mac’s notch.

🔗 Parse.bot — turn any website into an API.

Issue #32 of Computing Education Things was written while listening to:

🔗 Quick Links

🎧 Listen to Computing Education Things Podcast

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

Buy Me A Coffee

#31 — The Illusion of Competence

Daniel Prol — Fri, 06 Feb 2026 11:20:30 GMT

Image credit: Leap-N-Learn Book Club discussion (via LeapNLearn.com)

Mo Bitar

After two years of vibecoding, I'm back to writing by hand

Most people’s journey with AI coding starts the same: you give it a simple task. You’re impressed. So you give it a large task. You’re even more impressed…

4 months ago · 98 likes · 55 comments · Mo

Those of us who work in technology tend to live in a bubble. And to be honest, there are good reasons for that: the January 2026 Anthropic Economic Index shows that a large share of interactions is focused on programming and code. For those of us who are curious and love to learn, it’s hard to remember a more exciting moment than this. Yet understanding what AI is truly changing—and what it remains unable to touch—requires us to look outward, drawing lessons from other disciplines where tools have long augmented human skill.

There’s a noticeable gap in how people experience AI today. Some users—typically those with programming expertise—are embracing AI for efficiency, with real cultural impact on their teams, as Iñigo Medina has noted. Lower barriers to code mean more people can contribute, and teams that once had to ask for changes can now solve problems on their own. Others use AI constantly too, but mostly for incremental improvements—writing emails, summarizing documents, speeding up routine work. Helpful, yes, but transformative? Not quite.

At the same time, this changes how competence shows up. In The ABCs of How We Learn, Schwartz and Tsang define competence as “the feeling that one is capable of achieving desired goals and gaining mastery.” With AI, that feeling and the underlying capability can diverge. When AI-generated work is reviewed in small pieces—diffs, test results, isolated outputs—it often appears solid. Some developers have noted an additional risk: newer AI coding assistants can fail in ways that are hard to detect, producing code that runs and looks plausible while quietly doing the wrong thing—reinforcing the illusion of competence. The code works, tests pass, reviews approve. Over time, though, teams are discovering that understanding the whole still matters. Reading a system end to end, revisiting old code, or making deep changes requires more than “it worked at the time.” It requires someone who has kept a mental model of the whole.

This isn’t a failure of AI so much as a reminder of what it doesn’t provide. AI is very good at producing plausible results. It doesn’t maintain coherence over time—style, structure, long-term maintainability. Think naming conventions that drift, or architectural patterns that fragment across modules—unless a human actively enforces those qualities.

Seen this way, the so-called AI gap and the illusion of competence are part of the same learning curve. Early users extract more value not just from better prompts, but from recognizing when AI output needs human intervention—when to accept, when to revise, when to rebuild from scratch. A recent study published in Science finds that while AI has increased coding output overall, the gains accrue almost entirely to senior developers, with little measurable benefit for early-career users despite higher adoption rates. As organizations mature in their AI use, the challenge becomes balancing delegation with understanding—using AI to move faster without losing sight of how things fit together.

This tension is not new. As a developer recently observed, some of the most productive days involve writing almost no code at all. Time spent observing users, asking questions, and validating assumptions can create far more value than hours of uninterrupted coding—yet these forms of work are often invisible to the metrics we use to define developer productivity.

That perspective becomes clearer when we step outside the tech bubble and borrow ideas from other disciplines—medicine, architecture, skilled trades—where tools have long augmented human work without replacing responsibility or judgment. AI, it turns out, fits that pattern rather well.

The same dynamic is now playing out in education, where the costs of misplaced confidence can be even harder to reverse. The rapid adoption of AI in classrooms appears to be fueling a growing illusion of competence. Maybe students don’t need more shortcuts; they need safeguards against the deskilling that is already underway. Lower lecture attendance, more superficial reading, and heavy reliance on AI form a pattern that makes it easier to get by, but empties learning of its meaning. The cost is not only academic: those small moments of achievement—the “yes, I did it”—that once sustained motivation and pride in one’s work quietly disappear.

Paradoxically, even boring tasks play an important formative role. A deep qualitative study by researchers at Sciences Po suggests that automating them does not necessarily make work more engaging; in many cases, it does the opposite. When such tasks are handed off to LLMs, work can become more alienating and less meaningful. AI does not merely reveal something about our work—the way we use it can actively generate boredom. This is where the confidence trap emerges: we feel more productive because things move faster, while the experiences that build understanding, judgment, and motivation slowly erode.

A similar conclusion is echoed by developers who have pushed AI-assisted coding further than most. After two years of working extensively with agentic systems, Mo Bitar describes returning to writing code by hand—not out of nostalgia, but pure pragmatism. Once factors like code quality, long-term maintainability, and the cost of accumulating technical debt are taken into account, manual coding often turned out to be faster. For software engineers fatigued by vibecoding, his experience serves as a useful counterpoint:

Does EdTech work?

There is now growing evidence that education technology has not delivered on its promises, at least in K–12 settings. As Jared Horvath puts it in a recent The Economist article: “In nearly every context, ed tech doesn’t come close to the minimum threshold for meaningful learning impact.” The case of McPherson Middle School in Kansas follows a familiar script: laptops introduced with high hopes, growing concerns about learning impact, constant distraction, and teachers reduced to digital hall monitors. This is not an outlier. Independent meta-analyses—including a Stanford University review of 119 studies—find that the learning gains associated with ed-tech interventions are generally modest, highly variable, and strongly dependent on context, skill type, and implementation.

The persistence of the ed-tech narrative seems to have more to do with how these tools are promoted than with how they are typically used in practice. Even the most frequently cited “success stories” apply to a small minority of students. Usage data from platforms like Khan Academy show that most students barely engage, and that measurable gains are concentrated among roughly the top 10% of users—often those already motivated or supported at home. At the population level, average effects remain modest.

Taken together, this body of evidence—including the findings discussed above and the recent report from the Brookings Institution—should not be read as an argument against AI in education. Rather, it points to the need for greater care in how these tools are selected, implemented, and evaluated, and for investing just as seriously in the human conditions that make learning possible in the first place.

The thing that makes us human

This week I really enjoyed this episode from the Thoughtbot Podcast (episode 601), where the hosts (Chad Pytel and William Larry) dissect how experienced developers are actually using AI day to day. I agree with their case for maintaining ownership of the codebase and balancing coding enjoyment with AI through small increments, TDD, and treating AI as a pair programmer.

The conversation also surfaces the practical risks of relying too heavily on agents—hallucinated reinventions of built-in framework features, false confidence, and security or architectural gaps in AI-generated systems. Importantly, they don’t stop at the risks: they also discuss where AI genuinely shines, especially in UI prototyping and learning new stacks.

Beyond day-to-day practice, they grapple with broader implications as well: hype-cycle economics, ethics and energy costs, boundaries around creative work, and a likely future where capable local models reduce dependence on cloud LLMs.

Tools mentioned:

VS Code – Main editor used, especially for JavaScript-heavy workflows.
GitHub Copilot – Inline AI assistance inside VS Code.
Zed – New editor being tested, focused on performance and AI-native workflows.
Claude – Used via terminal and editor integrations for coding support.
ChatGPT – Mentioned as an interchangeable LLM option within editor workflows.
Supabase – Cited as a common default backend in AI-generated prototypes.
Lovable – Example of rapid prototyping tools used by non-technical founders.

Why cheating with AI is a question of virtues

This week I’ve been thinking about virtues and the real impact they have on our everyday decisions. A good example is cheating with AI. If we look at it through the lens of the virtue of justice—classically defined as giving each person what is due to them—several forms of injustice quickly emerge when we cheat with AI.

We are unjust to our classmates who are actually putting in the effort; we waste the instructor’s time; and, perhaps most importantly, we are unjust to ourselves, because we claim a level of learning or competence that we haven’t truly developed.

The Absent-Minded Professor

How Colleges Should Respond to the Intelligence Revolution

Thank you for being here. Please consider supporting my work by sharing my writing with a friend or taking out a paid subscription…

4 months ago · 15 likes · 3 comments · Josh Brake

This is where the analogy Josh Brake draws with the so-called cognitive revolution fits especially well. Just as the Green Revolution brought an abundance of food—making it easier to eat well, but also easier to eat poorly—AI today puts intelligence “on tap.” It has never been easier to produce text, solve problems, or generate code. That abundance opens up enormous possibilities, but also new risks.

In such a world, those who have cultivated the virtues needed to use these tools well can truly flourish intellectually. They will use AI as a means to go deeper, not as a way to avoid effort. Following the analogy, they are the ones who know how to choose a healthy cognitive diet. But when abundance is not accompanied by maturity and character, the result is a kind of cognitive intoxication: using the tool not to learn more, but to think less.

For this reason, the problem of cheating with AI is not merely technological or disciplinary; it is deeply formative. The underlying question is not “Can I get away with this?” but “What kind of person am I becoming when I systematically delegate my intellectual effort?”. Here, justice appears again—not only as external justice, but as internal justice: giving yourself what you truly need in order to grow.

This connects to a broader question about the role of the college. For years, higher education has been treated almost exclusively as a space for acquiring competitive skills. That is necessary, but clearly insufficient. In the context of ubiquitous AI, the formation of character—curiosity, intellectual honesty, humility, fortitude, responsibility—stops being a secondary goal and moves to the center.

Perhaps the key question for colleges today is not “How do we respond to AI?” but “How are we forming our students to live intellectually healthy and genuinely human lives in a world of cognitive abundance?”.

If you’d like to go deeper into these questions, you can apply—as I did—to ITiCSE WG10: Teamwork in Computing Education: Skills, Values, and Virtues. Membership application form #1 closes Monday, February 16, 2026 (AoE).

🔍 Resources for Learning CS

→ The World’s Most Important Machine

For over 50 years, transistors got smaller and chips got more powerful, doubling every two years. But around 2015, this progress hit a wall. This is the story of the $400 million machine that broke through: using extreme ultraviolet light, atomically smooth mirrors, and plasmas hotter than the sun, ASML built what most experts thought was impossible, enabling all modern advanced chips.

→ Speeding up NumPy with parallelism

A practical guide to speeding up NumPy: when threading helps, when Numba helps more, and why memory bandwidth caps your gains—plus why “automatic” parallelism can be risky in real code.

→ Faster.dev

Aaron Francis released an education platform focused on building software faster with AI.

→ Introduction to PostgreSQL Indexes

If you’ve ever added an index and been surprised that performance didn’t improve, this guide is worth your time. It explains how PostgreSQL stores data, when indexes help, and when they can make things worse—using real EXPLAIN output and concrete tradeoffs.

→ A BERTopic tutorial

This is a practical playbook for structuring large text corpora, covering data cleaning, model selection, parameter tuning, topic merging, and result validation through visualization and qualitative analysis.

🔍 Resources for Teaching CS

→ Crowdsourcing criteria for statistical infographics

This sheet is used for students to collaboratively propose rubric criteria for evaluating statistical infographics. The activity helps make expectations explicit, builds student ownership of assessment standards, and supports peer learning in a genre that is new to most students.

→ Josh Brake on teaching in Higher Ed

Fascinating conversation between Josh Brake and Bonni Stachowiak on the Teaching in Higher Ed podcast: AI as an “e-bike for the mind,” the risks of replacing effort, why intention, better questions, and accountability matter, practical tools for prototyping small web apps and games tailored to classroom needs (Claude and Gemini), and reading recs—Ursula Franklin and Vonnegut’s Player Piano—as lenses for thinking more clearly about what these tools are doing to us. The episode and show notes can be found here.

→ Stanford CS149 Parallel Computing

A solid Stanford course on parallel computing with a strong focus on real hardware–software trade-offs. Link to Course Playlist | Link to Course Slides.

→ Python for Data Analysis, 3E

The full text of O’Reilly’s Python for Data Analysis (3rd edition) is freely available on the author’s website.

→ Intro to Quantum Computing (Undergraduate)

A practical, funding-friendly path combines Qiskit (hands-on, browser-based labs), Microsoft’s Azure Quantum (self-paced programming intuition), Quirk (visual circuit intuition), and Scott Aaronson’s Quantum Computing Since Democritus for conceptual grounding—enough to make students quantum aware without hardware or heavy math.

🦄 Quick bytes from the industry

→ The creator of OpenClaw

The Austrian engineer Peter Steinberger became a legend in 2010 after releasing a PDF viewer—PSPDFKit—that outperformed Apple’s own. What began as a side project quickly grew into a company whose software now runs on more than a billion devices. After years of intense work, Steinberger burned out, stepped away from tech, and disappeared for a while.

Now he’s back—building in a radically different way. With OpenClaw, a personal assistant that can act directly on your computer, Steinberger works with AI agents not as tools, but as collaborators. He no longer reads most of the code he ships. Instead, he focuses on system architecture and how software feels, treating pull requests as prompt requests.

The key shift, he argues, is “closing the loop”: designing systems where AI can run, test, debug, and validate its own work. The result is faster development—and, paradoxically, better software.

Two takeaways:

If you want AI agents to produce high-quality work, you have to design systems where the agent can verify its own output.
Everything is now one good question away—but you need to know what to ask.

→ Ryan Olson’s Career Journey

In this interview, Ryan Olson reflects on his career journey from failing a Facebook interview to becoming one of the key product engineers behind Instagram Stories, later leaving big tech to found a more human social app, Retro. It’s a story about the limits of technical interviews, the power of small teams, strong product taste, and choosing craft over scale.

The limits of technical interviews

Ryan opens with a painful failure: a Facebook interview where nerves took over, he froze on a data structures question, and the interviewer ended the session early. That rejection stuck with him for years. Looking back, he sees it as proof that technical interviews often test the wrong things—especially when high performers can fail due to stress, timing, or arbitrary questions.

After that rejection, Ryan joined Flipboard as an intern, turning down an Amazon offer in the process. Flipboard had an unusually high density of talent—former Apple engineers, people who had built core iOS frameworks, and interns who would later go on to found companies like Figma and OpenSea. What stood out most was the mindset: many weren’t aiming for “a job,” but to start companies. Flipboard became his real education in iOS development and product thinking.

During that time, Ryan also built and open-sourced an iOS debugging tool that became widely used, even inside Facebook and Instagram. Ironically, interviewers still ignored it and focused on whiteboard problems. The real value of the project came indirectly: respected software engineers noticed his work and advocated for him behind the scenes.

Instagram

Ryan joined Instagram when the iOS team was small and unstable, but that chaos created opportunity. By profiling startup performance and fixing basic issues, he made changes that dramatically reduced crashes and improved launch time. Instagram’s early culture rewarded outsized impact, reinforcing his belief that small and focused teams can move fast and matter.

He later led a major visual redesign of Instagram (“Whiteout”), working closely with designers in tight feedback loops. While leadership initially encouraged shipping without A/B testing, the launch was delayed at the last minute to run experiments. Ryan’s takeaway was clear: experiments are great for optimization, but too much testing leads to incrementalism. Big product shifts require conviction and strong product taste.

That same philosophy shaped Instagram Stories. At the time, Instagram felt like it was losing everyday users to Snapchat. To address this, Ryan led the iOS effort for Stories with a tiny team. Ownership was clear, decisions were fast, and everyone was deeply invested. The work was intense and unsustainable long-term, but it resulted in a polished product that reshaped Instagram—and the broader social media landscape.

Stories succeeded in part because Instagram already had the social graph; it just lacked the right sharing format. The feature gave users a low-pressure way to post again. The team also obsessed over small interactions—like tap navigation and “hold to pause”—that later became standard. Ryan emphasizes that software engineers have real power: if something feels wrong, you can just build a better version.

Growth, promotions, and learning from failure

After shipping both the redesign and Stories, Ryan was promoted quickly. He notes that promotions are partly about performance, but also about incentives and timing. More important than titles was the freedom he earned to work on the most important projects.

Not all of those projects succeeded. Ryan later worked on IGTV, a bold attempt to bring long-form video to Instagram. Despite a strong team and a clear vision, it failed. Creators didn’t want to make long vertical videos, and AI attempts to reformat content were poorly received. The experience reinforced a key lesson: even great teams can miss when product–market fit isn’t there.

As Instagram continued to grow, Ryan felt the original product culture fading. In response, he started IG Labs—an “elite” team designed to explore new ideas outside the main org structure. Most experiments didn’t ship, but some features did, and the group helped keep experimentation and innovation alive inside a much larger company.

Leadership philosophy, leaving big tech and advice

Ryan also experimented with a hybrid role, managing people while continuing to code. He argues that senior engineers and managers should stay hands-on. Writing code forces better system design and greater responsibility for outcomes—what he describes as having “skin in the game.” While he admits this approach isn’t optimal for climbing career ladders, it aligned closely with his values.

Eventually, Ryan left big tech to start a product studio, with Retro as its first app—a social product focused only on real friends, with no followers, no algorithmic feed, and no ads. Growth is harder and monetization relies on subscriptions, but the team has seen what’s possible: Retro reached critical mass in Taiwan, where it briefly topped the App Store and continues to rank highly in its category. The goal isn’t maximum engagement, but building products that feel good to use—and good to leave.

His advice to new graduates is simple: learn the tools of your time and build things for people. Today, that means AI tools. New graduates have an advantage because they can adopt new tools without legacy thinking. If you can reason well and know when not to trust the tools, you’ll remain valuable.

🌎 Computing Education Community

Call for papers for an IEEE Software special issue on Human-Centric AI in Software Engineering.
Anastasiia Birillo shared resources on LinkedIn including this paper on AI-generated hints, a GenAI teaching tools demo (see you there!), and a talk on interactive teaching with Kotlin Notebooks.
A free, full-day Professional Development Session at SIGCSE 2026 for new and aspiring CS educators, covering career paths, teaching, the academic job search, AI in education, and work–life balance. Speakers include Colleen Lewis and Leo Porter.
CSEE&T 2026 has extended the submission deadline for full and short research papers to March 1, 2026, with abstracts for full papers due February 20.
A Friday evening tutorial at SIGCSE TS 2026 on using specifications grading to create more equitable, outcome-focused assessments, while also reducing grading disputes and workload.
DUBOIS Data Science Symposium (May 18–20, 2026, Auburn).
Upcoming deadlines for Consortium for Computing Sciences in Colleges – Central Plains: student hackathon (reg. Feb 6), posters/papers (Mar 27), and programming contest (Apr 11 at Drury University).
If you teach digital logic, computer architecture, FPGAs, or RISC-V, consider Tutorial 104 at SIGCSE 2026 on development containers and accessible, hands-on assignments.
The Raspberry Pi Foundation hosts its next research seminar on Feb 10 (17:00–18:30 GMT), featuring Thema Monroe-White (George Mason University) on race-conscious algorithmic approaches to AI, examining intersectional bias in LLMs and implications for teaching and education research.
IntroCS-POGIL is an NSF-funded project that promotes the use of guided-inquiry learning in introductory computer science, helping students build core concepts and skills through structured, team-based activities.
Just a quick reminder about this SIGCSE-affiliated event on Team-Based Software Projects!

🤔 Thought(s) For You to Ponder…

When we interpret everything from our own perspective, we tend to misread things because of our limited view (we lack objectivity). That’s why we all benefit from someone who can look at things from the outside and guide us in a tailored way. Thanks for making me think, Josh Brake!

I had never heard of the concept of Agent Experience (AX) before, but it seems to be becoming increasingly relevant. The author, one of the founders of Netlify, states: “Most of the products or platforms we build will become irrelevant unless autonomous agents can use them efficiently on behalf of their users.”

It’s good to have at least one hobby that is purely manual—something you do with your hands. There’s growing evidence of the deep connection between hand and mind, which helps explain why practices like knitting, ceramics, woodworking, or cooking are being rediscovered. Not as obligations, but as creative, restorative acts. Aristotle believed that alongside intellectual work, every free person should know a manual craft—because it develops dimensions of the human being that thinking alone cannot. In a world saturated with screens, working with real materials reconnects us with reality in a deeper, more truthful way. As Julie Kilcur quietly reminds us:

Even if your day-to-day tasks lend themselves more to the cerebral, doing physical work is important to our souls as we fight an increasingly contested reality that we are wholly embodied creatures who exist in the material world with bodies that are meant for loving and clapping and cleaning up and clinking glasses and climbing mountains and swimming in the ocean and ladling soup into bowls and digging in the dirt and holding wriggling children in one hand and a well-worn read aloud in the other.

Liturgy of Ours

Live a Quiet Life and Work With Your Hands

Spoiler alert: There are a few minor spoilers in this piece—though I tried to keep it as friendly to the new viewer as possible…

4 months ago · 257 likes · 35 comments · Julie Kilcur

We are living in the era of the individual creator.

It’s less than 1% of the papers, but as reviewers, we need to be more vigilant about AI slop to prevent papers with hallucinated citations from slipping through and undermining trust in academic conferences. The presence of these kinds of errors in work that has gone through peer review can’t be overlooked.

I’m glad to see that more and more full degree programs are being offered in English in Spain. It’s a trend I expect will continue to grow.

Laracasts cut 40% of its workforce. See the tweet for the comments.

You must’ve been living under a rock this past week if you haven’t seen the OpenClaw hype: a local, agentic assistant that runs on your own machine and can act on files and apps via MCP and skills. A big part of what made it go viral is that you can talk to it from WhatsApp or Telegram—almost like having your computer on call. The catch: it doesn’t include the model (you usually pay API costs separately), it’s not really for non-technical users yet, and the security tradeoffs are real if you’re giving an agent access to your browser, files, and accounts. Simon Willison dives deep into the so-called “Facebook for AI agents” Moltbook (a social network where only AI assistants interact), in this detailed write-up.

You may disagree with Iñaki Gabilondo on many things—and I probably do too—but it’s a real pleasure to listen to him talk about Spanish broadcasting. This episode with Guille had several moments that really stayed with me.

One of them is a simple reminder: it’s worth listening very carefully to recommendations from people who truly know their craft.

But the analogy that struck me most was this one:

When someone spends six or seven hours communicating with listeners every day, the best thing they can do is stop trying to wear a disguise. It’s like a teacher who walks into a classroom of forty students every day. You might manage to put on a mask for a single day—but over the course of a school year, the best option is to be who you really are. Try to become the person you want to be, because it’s inevitable that this is what listeners will ultimately perceive. There are too many hours, too many days, too many years to pretend to be something else.

And finally, this reflection on responsibility really landed with me:

Even if the reasons I had for being angry were still valid to me, my way of expressing that anger would not be. Anyone with responsibility behind a major microphone has an obligation to be extremely careful with how they express their views—not just what they think, but how they say it. Because the way you express disagreement doesn’t just convey an opinion; it also teaches others how to disagree, how to live alongside difference.

Two good peers from my time in LatAm recently had a really interesting conversation in Spanish—and Santiago Zavala made a really insightful point:

Even if AI makes building software much easier, creating a legendary product is still hard. Users still need real solutions, and there’s still a lot left to explore in terms of what the best version of many tools looks like. The point is that “simple” products can still be improved, and we haven’t yet found the best version of everything.
What these new tools change is the scope of what’s possible: we’ll be able to build much more software. But that leads to two consequences. First, users will have far more options, which raises the bar—only the best will stand out. Second, distribution becomes even more important. With so many alternatives, most products will go unnoticed, so earning attention matters more than ever.

📌 Research Corner

I’m really enjoying the Research Methods course. Our professor, Omprakash Gnawali, has academic experience at Stanford, USC, and MIT, and the course is highly practical, with plenty of hands-on guidance and actionable advice. If you’re doing a PhD or supervising students, you can share the publicly available slides with them.

Teaching gets better when it’s done collaboratively, and observing others teach is one of the most powerful forms of professional development. I’m excited to be a TA for Software Design this semester under the supervision of my advisor, Amin. I can’t think of a more relevant class for the times we’re living in—where judgment, taste, responsibility, and domain knowledge remain irreplaceable. We closely follow the software construction course from Carnegie Mellon in case you’re interested in their materials. We also use CodeCheck to create in-class quizzes, and occasionally for attendance and participation.

I’ve generally been pro-flexibility when it comes to remote work. I like having the best of both worlds. But one thing has always been clear to me: when I came to campus, I saw people, I socialized, and conversations happened. From home, I simply didn’t have those casual interactions—I didn’t stumble into ideas or topics I wasn’t actively seeking out.

A good example was a hangout I had this week with my PhD friend Germán, who works in a field—cybersecurity—that feels like a completely different language to me. Every time we talk, I learn a ton. Those kinds of serendipitous conversations are much harder to replicate remotely.

🪁 Leisure Line

After five months in Houston without touching one, this tortilla tasted absolutely incredible 😋

The following day, the culinary journey took a different turn. Lunch with lab mates turned into a deep dive into authentic Arabic cuisine at Hadramout Restaurant.

To finish on a sweet note, a nighttime discovery on Westheimer: a Syrian chocolate shop well worth the stop, Chocolate Zeina.

📖📺🍿 Currently Reading, Watching, Listening

I’ve been a big fan of Samuel Arbesman ever since I started reading his Substack newsletter, Cabinet of Wonders. I recently picked up his latest book, The Magic of Code. I’m just cracking it open, but even in the first few pages, there’s already a lot to think about.

Issue #31 of Computing Education Things was written while listening to:

🌐 Cool things from around the internet

🔗 Logosystem — the biggest logo design library for inspiration.

🔗 OpenCode — an open source AI coding agent.

🔗 Isometric NYC — an isometric map of NYC—absolutely fascinating.

🔗 Data To Art — this project showcases works that transform datasets into art.

🔗 Quick Links

🎧 Listen to Computing Education Things Podcast

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

Buy Me A Coffee

#30 — The fight to keep thinking

Daniel Prol — Fri, 30 Jan 2026 11:20:58 GMT

This edition of Derek Thompson’s newsletter argues that the greatest risk of AI is that people may stop thinking. Beyond the common argument that students and professionals are outsourcing writing and reasoning to models—thus eroding basic cognitive skills—Thompson contends that writing is thinking, and deep reading is training for the mind.

I believe the rise of LLMs is further accelerating the shift away from long-form reading. We now tend to prefer other formats, such as podcasts, videos or social media, and our content consumption has become more fragmented.

Yet in this new economy, we still need the ability for symbolic and systemic thinking. As both Cal Newport and Walter Ong have noted, reading and writing are the serum that develops deep thought.

The conclusion: we must sit with difficult ideas, read, write without shortcuts, and resist the comfort of total automation to prevent the decline of thinking people.

Derek Thompson

The End of Thinking

This is an expanded and revised version of an essay that originally ran in The Argument, an online magazine where I am a contributing writer…

8 months ago · 1635 likes · 120 comments · Derek Thompson

Example 1: Hussein Nasser argues that the most effective way to design software is by writing, not coding or diagramming first. His process starts with detailed workflows, moves to a technical design overview, then component-level documents, and only ends with diagrams—accepting the time cost in exchange for clarity, rigor, and confidence when reasoning about the system.

Hussein Nasser

How I design software

How I design software…

6 months ago · 119 likes · Hussein Nasser

Example 2: Windsurf Codemaps is an AI-powered way to build structured, navigable mental models of large codebases. Instead of replacing thinking with “vibe coding,” Codemaps helps engineers read, reason about, and stay accountable for complex systems—turning AI into a tool that keeps your brain on, not off.

Brute Force Exploration

Ralph Wiggum is a script that runs Claude Code in an infinite loop, feeding it prompts until it either solves a problem or runs out of budget. It relates to today’s topic because this new methodology involves no guided exploration—just a loop that generates, evaluates, fails, and repeats. The thinking isn’t in the process itself but in the acceptance criteria. It’ll probably work; after all, it’s still a valid approach, much like machine learning or many other areas of computer science.

Outside of critical systems—where the priority may still be on hand-coding—the focus in software engineering is increasingly shifting from writing software by hand to understanding what to build and why. It’s about exercising judgment, taste, and strategic thinking. It’s about tackling complex problems, designing robust solutions, and making tough calls. In that context, AI is just another tool.

As Eduardo Díaz puts it on his blog, we are complexity managers—translators between the chaotic world of human problems and the formal world of computational instructions.

What worries me is that Ralph Wiggum doesn’t actually think. He says things that sound accidentally profound—but he doesn’t understand why they’re profound. I believe that kind of thinking is still our responsibility—at least for as long as we have the judgment to know what to build and the taste to recognize when it’s well done.

If Ralph Wiggum feels unsettling, Michael Arnaldi’s “The Death of Software Development” pushes the same idea further. His argument is simple: software development as a craft is dead, but software engineering is alive and well. What matters now is not writing code, but designing the processes that produce it. With the right methods, “good enough” agents can build systems that once required entire teams:

The role has transformed. Engineers are no longer writing software — they’re designing higher-order systems. They’ve moved from crafting code to designing systems that write code. They build techniques. They build skills. They develop the mental models and architectural intuitions that guide AI toward good solutions. This new reality requires rethinking everything. Forty years of best practices are now outdated. The patterns we relied on, the team structures we built, the processes we followed — all of it needs to be reconsidered. Individuals are far more powerful than before. A single person with the right skills can now do what used to require an entire team.

If Arnaldi describes the technical inflection point, Peter Rojas’s “How Mozilla Builds Now” articulates the human response. The real work is making systems legible, bounded, and worthy of trust. Rather than optimizing for speed alone, Mozilla frames building as a responsibility — designing tools that assume users can think, judge, and remain in control.

Long live the factory

Quinn and Thorsten recorded a new episode of Raising an Agent, many months after the last one. The agent’s all grown up now—and hey: “the assistant’s dead, long live the factory.”

FYI: I created an AI assistant last year 😂 There are still ways to capitalize on it (e.g., research manifestos, proposals, etc.).

This week, I saw a thought-provoking tweet from Ryan Dahl, creator of Node.js and Deno:

This has been said a thousand times before, but allow me to add my own voice: the era of humans writing code is over. Disturbing for those of us who identify as SWEs, but no less true. That’s not to say SWEs don’t have work to do, but writing syntax directly is not it.

For years, the dominant mental model for AI in software engineering has been the assistant: a one-to-one interaction, a conversational partner that helps you write code faster, autocomplete functions, or clean up rough edges. Useful, yes—but fundamentally limited.

What Thorsten Ball articulates in this conversation feels like something deeper: the assistant is no longer the center of gravity. The factory is.

This shift didn’t happen in a vacuum. It coincides with a real inflection point in model capabilities. When Gemini 3 landed, it quietly reset expectations. A week later, Sonnet 4.5 (and its peers at the frontier) erased what many thought were hard limits: agents couldn’t sustain long tasks or operate autonomously without constant human steering. Suddenly, those assumptions were outdated.

What’s striking is not just that these models can write more code—it’s that they can hold the shape of a problem for longer. They don’t need to be micromanaged step by step. You can give them a longer leash—and they don’t just run, they navigate.

That’s where the “factory” metaphor becomes especially compelling. If agents writing code is now a given, the more interesting question is: what happens when we stop treating them as tools waiting for instructions, and instead see them as systems we can embed directly into our codebases—systems that can explore, test, verify, and iterate independently?

This reframes the human role entirely. The challenge is no longer writing every line of code correctly—it’s preserving the thinking behind it: choosing the right problems to solve, designing environments where agents can thrive, and building feedback loops that encode judgment, not just execution.

There’s a quiet but profound tension here. As agents get better, the temptation is to think less—to delegate not only labor, but thinking itself. Yet the opposite may be required. To build agent-native systems, humans must think more clearly about structure, constraints, verification, and intent. Deep thought doesn’t disappear; it moves upstream. In that sense, agentic factories aren’t the end of thinking—they may be the catalyst that compels us to sharpen it.

🔍 Resources for Learning CS

→ What is Continuous Delivery?

In this post, Thierry de Pauw emphasizes that Continuous Delivery isn’t an end state or a one-time project—it’s an ongoing practice of continuous improvement aimed at making deployments safe, reliable, and routine.

→ Using enumerate and List Transformations in Python

This is a great video by Prof. Kurmas featuring examples of how to use Python’s enumerate function and how to build a new list based on an existing one.

→ Book recommendation

The Magic of Code by Samuel Arbesman is a valuable supplementary resource for computer science education. Framing computing as a humanistic liberal art, the book explores the cultural, historical, and philosophical dimensions of code and computation. While it does not teach programming directly, it offers rich insights that can deepen one’s understanding of computing as a discipline. A PDF of the introduction is available here for those interested.

→ Deep Learning with Python, Third Edition

A clear & code-first guide to deep learning—from first principles to generative models, by François Chollet and Matthew Watson. No heavy math, just hands-on examples in Keras, PyTorch, TensorFlow, and JAX. Read online.

→ Large Language Model tools for R

Explore the R packages that turn large language models into practical tools—from structured prompts and local model support to IDE assistants and agents that work directly in your R session. This guide is also available in Spanish.

→ A Diary of a Data Engineer

Thinking about becoming a data engineer? This honest post by Simon Späti captures what the job really feels like—beyond the tools and hype.

→ Pioneers of RL: Barto & Sutton

ByteCast Episode 80 features Turing Award winners Andrew Barto and Richard Sutton, pioneers of reinforcement learning. Hear them discuss the foundations of modern AI—including AlphaGo and ChatGPT.

Don’t chase fashionable problems—follow your intrinsic curiosity and look for the “obvious” truths your field is overlooking, because lasting breakthroughs come from developing what others don’t yet see, not from riding the trend of the moment.

🔍 Resources for Teaching CS

→ Developing Syllabus Statements for AI

If you’re still working on your syllabus, you might find some inspiration here: how to craft AI policies that go beyond simply allowing or banning its use. The goal isn’t control, but helping students understand why certain uses of AI can support—or undermine—their learning, and guiding them to develop sound judgment.

→ Computer Science In 100 Images

I highly recommend Computer Science in 100 Images by Dr. Ashish Bamania. The book distills 100 foundational computer science and software engineering concepts into clear, thoughtfully designed visuals, each accompanied by concise explanations.

→ Intro to Bias and LLMs

This video from Google does a great job explaining bias in models and could serve as a quick primer for students to get a sense of what bias can look like. It’s also a useful starting point for discussing more serious forms of bias that may arise. This second video, focused on LLMs, offers a solid overview of how large language models are trained, how they work, and even touches on transformers. It gets a bit technical toward the end, but overall, it’s a great introduction. Thanks to Ismael for sharing these valuable resources!

→ An episode on rethinking Learning Objectives with AI

This episode did a great job of exploring the nuance in the tension between acknowledging that using AI may now be a learning objective, while also recognizing that it might be one of the first objectives potentially at odds with others. However, this means we need to rethink our learning objectives and how we assess them — which is possible, though challenging.

→ Teaching Neural Networks with Snap!

Check out Jens Mönig, research expert at Snap!, as he shows how this visual programming language—developed by SAP in collaboration with UC Berkeley—can be used to create, train, and understand neural networks in a fun and accessible way. If you’re interested in diving deeper into Snap!, there are some great resources and examples worth checking out:

Mark Guzdial’s course using Snap!: Alien Anatomy: How ChatGPT Works
SAP blog post: What’s Next: Generative AI with Snap! GPT
Ken Kahn’s project collection: AI tools for beginners
Eckart Modrow’s SciSnap: Programming with SciSnap PDF

→ Building intuition about LLMs

I really enjoyed Mark Riedl’s blog posts—they’re some of the best at building intuition around how LLMs work. For those of you teaching or designing AI-related curriculum, his writing strikes a great balance: technically grounded, yet accessible enough to spark classroom discussion or help students connect theory to practice.

I especially recommend these two posts:

They’re valuable reads for anyone thinking about how to teach LLMs in a way that moves beyond architecture diagrams and gets to the why behind model behavior.

→ AI & ML Materials for High School and Intro CS

Dr. Daniel Bauer (Columbia University) designed and taught a course on AI and ML for advanced high school students, blending core concepts with accessible materials. He’s made the course resources available on GitHub, some of which he also adapted for his introductory computing course at the university level.

David Czechowski uses unplugged AI activities in AP CS Principles to teach Machine Learning and Generative AI through hands-on/low-tech models. Students play Spicy Marshmallow, a reinforcement learning game inspired by MENACE (demo, write-up, middle-school version), and explore generative AI with a Markov Chain built from magnetic poetry (DIY kit, Scratch demo).

→ Causal Inference in Practice

Two hands-on courses—Program Evaluation for Public Service (Dr. Andrew Heiss, Georgia State University) and Causal Inference (Dr. Molly Offer-Westort, University of Wisconsin–Madison)—offer R-based, real-world approaches to teaching causality, research design, and ethical data analysis. Ideal for CS instructors seeking interdisciplinary examples of applied statistics.

🦄 Quick bytes from the industry

→ How Todoist is Built

Gonçalo Silva, CTO of Doist (the company behind Todoist and Twist), says: “I do see it as my responsibility as a leader… to help people navigate into the new world.” This idea closely aligns with what I discussed in Edition 17, about the importance of teaching others how to navigate an over-informed world. Silva also emphasizes that in the age of AI, clarity of thought—as well as the ability to communicate, plan, review, and iterate—is more crucial than ever. He stresses that the fundamentals remain irreplaceable when it comes to building that clarity and effectively guiding AI. Along the same lines, he points out that “architecture has always been important, but now even more so,” because producing more code, faster, makes mistakes significantly more costly—an observation that ties directly to Edition 22, which was inspired by ideas from Chris Lattner. He also notes a practical shift in day-to-day work—“more PRs instead of feature requests”—and highlights something essential for technical leadership today: as long as there’s critical thinking, it doesn’t matter whether someone is optimistic or skeptical about AI; both perspectives should be valued and rewarded.

Refactoring

How Todoist is Built ✅ — with Gonçalo Silva

Listen now…

Listen now

4 months ago · 24 likes · Luca Rossi

→ Nimit Sohoni on AI Research vs. Quant Careers

Nimit Sohoni (Stanford PhD) moved from AI research into quantitative research at Citadel Securities, and later joined Cartesia as an AI Research Scientist, where he works on real-time voice AI. In this conversation, he compares the day-to-day realities of AI research and quant work, explains what a PhD does unlock, and shares a practical playbook for building a career around deep technical fundamentals.

Five ideas worth keeping from the episode:

1. What a PhD Changes

Very few careers are strictly unavailable without a PhD (academia being the main exception), but a PhD makes certain paths much easier—especially industry AI research and quantitative finance. It helps both at the filtering stage (getting interviews) and in developing a key internal skill set: research taste. Without a PhD, AI roles often skew toward engineering-heavy work (infrastructure, data pipelines, evaluation), while a PhD can enable more exploratory, architecture-level research with longer time horizons. Still, both paths can succeed; startups in particular can allow role transitions if there is clear evidence of research ability. The core requirement is not the credential itself, but demonstrable depth in fundamentals.

2. Research Taste

Nimit emphasizes that “90% of the battle” in research is choosing the right problems—ones that are meaningful, tractable, and worth caring about. This skill develops through immersion: skimming large volumes of papers (X, coworkers), recognizing patterns in the literature, and progressively moving from small extensions to larger original ideas. Across both AI research and quant work, strong fundamentals matter more than domain-specific knowledge. Math intuition, numerical methods, linear algebra, stochastic calculus, and coding skill are transferable; finance concepts or new AI subfields can be learned on top. His advice is disciplined focus: build deep technical skills, read extensively, and avoid diluting your efforts across too many directions.

3. AI Research vs. Quant Finance

Quant and AI research share a similar math+CS core but differ sharply in culture. Quant roles vary widely (alpha generation, risk, data analysis, monetization), often involve heavy coding (frequently C++), and can offer surprisingly good work-life balance due to clustering around trading hours. Finance is also far more secretive than tech, even internally, with pod structures, strict information boundaries, non-competes, and garden leave designed to protect alpha. Compensation in quant is opaque and less standardized than tech, often bonus- and performance-driven, with higher upside and risk. In both quant and AI, top performers separate themselves through a mix of raw technical ability, execution speed, judgment, and being in the right place at the right time—but quant tends to have clearer, harder metrics for success. In this context, Nimit references firms such as Citadel Securities, Jane Street, Renaissance Technologies, Jump Trading, Hudson River Trading, as well as newer or more specialized players like XTX Markets, Radix Trading, and TGS Management.

4. Startups

After Citadel, Nimit returned to AI by joining Cartesia, drawn by the rapid acceleration of the field post–ChatGPT, trust in the founding team, and the opportunity to take risk after gaining stability. Cartesia focuses on real-time voice AI (text-to-speech, speech-to-text, voice agents), where low latency and tightly integrated end-to-end systems are critical for natural, usable interactions. He contrasts big AI labs—rich in resources but sometimes risk-averse—with startups that can challenge orthodoxy (e.g., state space models, hybrid architectures, learning tokenization from raw text). Voice, as a modality, benefits from architectural compression (SSMs) due to redundancy, making hybrid approaches especially powerful. More broadly, he argues that durable AI companies need both product and research: research-only startups face extreme risk, while products built purely on top of others’ models are easily overtaken as base models improve. Shipping real products reveals real failure modes, and those failures should guide model-level research.

5. Advice for Moving into AI Research

Nimit’s advice centers on depth over positioning. He argues that the most reliable path into AI research is to build strong fundamentals—coding ability, mathematical intuition, and the habit of reading papers—because solid technical foundations eventually attract opportunities on their own. For software engineers looking to pivot, credentials like a master’s degree can help signal readiness, but evidence of real skill matters more than titles. Startups can be especially effective environments for transitioning into research roles, since they allow more flexibility and organic role evolution—provided you can demonstrate the capability.

→ OpenAI and Codex with Thibault Sottiaux and Ed Bayes

Kevin Ball interviewed Thibault Sottiaux, Engineering Lead for Codex, and Ed Bayes, Codex Product Designer, both from OpenAI. Their conversation covered a range of topics, including models and harnesses, the future of multi-agent systems, model specialization, and considerations around latency and performance.

🌎 Computing Education Community

ByteByteGo is hiring for two roles: Technical Deep Dive Writer (System Design / AI Systems) and Lead Instructor (Building Production AI Systems). Both are part-time, remote roles focused on high-quality technical education and real-world AI systems. Apply at jobs@bytebytego.com.
Latent Space is hiring a Researcher/Writer to produce high-quality essays on AI research, products, and industry trends for a 100k+ audience. Remote (US timezone) or SF-based, $100k–$250k+ (part-time available).
LAK26 registration is now open and the full program is available.
RIT’s Future Faculty Career Exploration Program (Sept. 30–Oct. 3, 2026) is a four-day immersive experience for those interested in academic careers. The program offers insight into faculty life at RIT, including research, teaching, and service, along with networking, research presentations, job-talk feedback, and professional development. Travel, lodging, and meals may be covered for selected participants. The application deadline is February 27.
The first SIGCSE Journal Club of 2026 will take place on Monday, February 2 at 2:00 PM (UK time) and will be held online via Zoom. The session will focus on Semantic Waves in Computer Science Education and will feature Jane Waite (University of Cambridge) and Paul Curzon (Queen Mary University of London). The Zoom link is public (https://zoom.us/j/96465296256), while the password is available in the SIGCSE Slack channel, which can be joined via the instructions at sigcse.cs.manchester.ac.uk/join-us I’m in!
Working Groups are now forming for ITiCSE 2026, to be held in Madrid, Spain (July 10–15, 2026), and are seeking participants. WG6 focuses on building a global computing terminology resource by gathering and analyzing region-specific terms. WG4 examines how replication studies are conducted and published in computing education research, while WG10 explores teamwork in computing education, with attention to the skills, values, and virtues that enable effective collaboration. Find all Working Groups and application details here.
Heading to SIGCSE TS? Don’t miss Tutorial #405, Developing Peer Mentoring Programs for K–12 CS Teachers. This session will offer practical tools, strategies, and insights from four years of equity-focused mentoring work to support and sustain computer science teacher communities. You can also join the Professional Development Session for New and Aspiring Educators.
Undergraduate CS students are invited to submit abstracts for the CCSCNE 2026 Poster Competition at Smith College—top posters win cash prizes.
Haverford College is hiring a full-time Visiting Assistant Professor of Computer Science.
California State University, Long Beach is hiring three tenure-track Assistant Professors in Software Engineering, Computer Engineering, and Cybersecurity starting August 2026.
CS educators are invited to Devnexus 2026—a leading AI and Java conference—with a 40% discount using code EDUDN26F@CULTY. Join the educator session on March 5 and explore how Devnexus connects academia and industry.
The submission deadline for WCCCE 2026 is approaching.
CMMRS 2026 is a one-week pre-doctoral research school (Aug 3–7, Saarbrücken, Germany) where top undergrad and Master’s students explore CS with leading researchers from Cornell, Maryland, and Max Planck—apply by Feb 14, 2026 (AOE).
A new free online seminar series starting February 2026 will explore how AI is used and taught across disciplines beyond computing with a focus on computational literacy, critical evaluation, and classroom-ready research insights. Now available: a recent session featuring Jesús Moreno-León, where he explored methods for measuring and evaluating AI literacy, presenting an assessment instrument validated across multiple studies with thousands of primary and secondary students in Spain, and discussing its strengths and limitations.
USF’s Bellini College of AI, Cybersecurity and Computing is hiring Assistant and Associate Professors of Instruction for full-time instructional roles in AI, cybersecurity, and computer science, starting Fall 2026.

🤔 Thought(s) For You to Ponder…

I truly believe that the things which serve no practical purpose are what keep us from suffocating. It’s the useless that turns a flat life into one that flows—dynamic and driven by curiosity about the human spirit and the world around us. This thought was inspired by Sean Goedecke’s recent post.

Quan Nguyen on the soft skills gaining momentum for software engineers in 2026: Clearly defining problems, communicating effectively (especially in writing), asking sharp questions, and surfacing hidden assumptions or trade-offs.

It’s interesting how investing more, but with intention, in fewer well-made, high-quality clothes connects so naturally with the idea of sobriety:

Letters from the Old World

You Should Spend More on Clothes

The world is set up to ensure you continue to consume beyond your needs. In many sectors, the very economic model depends on continuous, unsustainable levels of product churn. This poses a problem for Christians in particular, who are warned against the love of wealth and material things…

4 months ago · 91 likes · 17 comments · Evan Amato and PJ Poscimur

I’m all for what Stacey Margarita Johnson is proposing in her new blog series, “Teaching the Good Stuff”: bringing current events, industry news, and real-world cases into the classroom so students can connect what they’re learning with what’s actually happening in the world. I think for that to happen, instructors need to be curious themselves first—but it’s just as important to create spaces where students can take the lead too (like a forum on Canvas, a Teams group, etc.).

I like the analogy brought up by Javier Vidal-Quadras about loving in the background—like a thought that’s always there, subtly noticeable, working beneath the surface, running in the unconscious until it becomes something that permeates every part of our life. In the background, yet always present.

The theme of fighting slop really resonates as generative AI increasingly strips authenticity from creation. It reflects a growing concern about not knowing whether code, content, or conversations are genuinely human or AI-generated.

Great piece. In what ways can I show love to others, especially those around me?

McKinsey is taking agents to the next level. It’s getting more embedded in the culture.

Back in my startup days, I always liked standup meetings—when they’re done properly. I don’t run them anymore as a researcher, but if I were to introduce one, I’d start by sharing this piece by Marc G. Gauthier.

Speaking of startups, Lawrence‘s honesty is admirable. He left a comfortable position at Amazon for a high-paying startup job in San Francisco, lured by the money, the fast pace, and the promise of making an impact. But at the startup, he quickly realized the culture demanded total commitment and emotional investment—something he didn’t feel. Within three weeks, he was let go after admitting he wasn’t enjoying the work. The experience turned out to be a tough but valuable lesson about the intensity of startup life, the importance of alignment, and choosing work that truly resonates—not just financially, but personally.

Lawrence's Substack

I left FAANG for a startup and regretted it

When I got the call from the CEO, I picked up the phone and gripped it, my body freezing up in anticipation…

5 months ago · 344 likes · 59 comments · Lawrence Tang

Although this MIT Media Lab paper came out last year, it still feels incredibly relevant. It tries to measure what happens in the brain when people write a text on their own, with the help of a search engine, or using ChatGPT. What they found is pretty clear: when people write without any assistance, their brains show more activity and stronger connections between different regions. When they use a search engine, the activity drops a bit. But when they use a model like ChatGPT, brain connectivity is the lowest of the three. This is what we call cognitive offloading. The more we outsource mental effort to a tool, the less our brain engages in the task. It’s kind of obvious, I know—but now there’s scientific evidence backing it up. What’s more, the study shows that the impact isn’t just “in the moment.” After multiple sessions, participants who used ChatGPT were less engaged, relied more on copy-pasting, and felt less ownership over what they wrote. They also had a harder time remembering or citing what they had written.

After weeks of having it on my watchlist, I finally caught up with the ACM-SIGSOFT session titled What Do Professional Software Developers Need to Know to Succeed in an Age of Artificial Intelligence?, featuring Matthew Kam (Google), Miaoxin Wang (Google), Vikram Tiwari (ClickUp), and moderated by Sridhar Chimalakonda and Andrew Begel. The talk is based on a recent paper that immediately clicked with this week’s theme. My main takeaway from the paper:

Productivity gains favor senior developers with strong mental models, while junior developers face a real risk of stagnation or deskilling as AI automates tasks that traditionally supported learning.

Also this week, I watched another ACM Tech Talk on How to Extract Meaningful Insights from Data. Angelica Lo Duca (IIT-CNR, University of Pisa) explored how insights emerge before any narrative or visualization, with Victor Yocco (ServiceNow) hosting the session. The talk emphasized practical analytical heuristics—temporal, spatial, multi-category, zoom, and anomaly analysis—to help decide what actually matters in real datasets. A recurring message was to focus on one insight at a time, treat peaks as outcomes rather than explanations, and recognize that flat data, missing values, and apparent anomalies often contain signal. The closing idea tied it all together: insights are relatively stable, but communication changes by audience.

📌 Research Corner

I’m currently deep in paper-writing mode. This week I’ve been testing out Flow for voice writing. It’s easy to use—just place your cursor, hit the keyboard shortcut, and start talking. It captures everything with impressive accuracy and even formats the text for you (perfect for creating lists). The free plan gives you 2,000 words per week. Available for Mac, iOS, and Windows.

Speaking of tools for researchers, TurboScribe is one of my favorite apps for transcribing audio. It’s fast, detects how many speakers are talking, and even lets you transcribe three 30-minute files for free every day. I use the paid version so I can get unlimited transcriptions and longer durations.

The deadline to submit the abstract for the EDM conference is next Monday. It’s been months of research—a whole experience. I know the bar is high, and this is my first paper as first author, but I’ll be happy as long as it makes an impact in the Computing Education Research community and the reviewers find it novel.

Speaking of conferences, I’ll be at SIGCSE TS from Wednesday, February 18 to Saturday, February 21. If you’re around and want to talk Computing Education, AI in programming, or anything related, I’d love to connect!

Attended a seminar on Wednesday where Hanjia Lyu (University of Rochester, Snap) went beyond high-level claims about multimodal and graph-based AI, grounding them in concrete case studies on data quality and alignment. What stood out was the emphasis on informative and selective data: from augmenting sparse video captions with language-model analysis in large-scale recommender systems, to human-in-the-loop filtering of vaccine-related tweets that dramatically reduced irrelevant data, and confidence-aware GNNs that adaptively handle heterophily in networked predictions. The talk closed with a sharp discussion on cultural misalignment in multimodal models (e.g., Simplified vs. Traditional Chinese), reinforcing a central theme: better human-aligned AI often comes not from more data, but from identifying the right data and understanding which human groups our models actually represent. I used Granola to record the talk while also taking my own notes. Afterwards, I wrote the copy you just read to expand and improve on those notes.

🪁 Leisure Line

The lack of ideas in this section is a clear sign that conference deadlines are approaching. Lots of routine work (blessed, nonetheless!). Wednesday was the feast of Saint Thomas Aquinas, and I’d like to mention that Pope Paul VI called him Lumen Ecclesiae—the Light of the Church. I think he’s a wonderful example of what it means to teach those who do not know, which is, in fact, a work of mercy. As educators, we have the opportunity to teach those who are unaware—and that’s a beautiful way to practice charity.

I recently listened to a podcast that explores his Summa. The episodes are short and definitely worth your time.

📖📺🍿 Currently Reading, Watching, Listening

This week, I watched Triumph of the Heart. It tells the powerful story of Saint Maximilian Kolbe during his imprisonment at Auschwitz. The film is available for $20 on this website—and it’s worth it.

I also watched American Sniper. A movie that does a great job portraying PTSD and its consequences. A devastating ending that shows the horrors of war. I really liked it.

A 40-minute session with James Baxter, where he explains how to animate walking cycles. This is essential viewing for anyone interested in learning animation or curious to see how a professional animator approaches walks.

Last Friday, Valentino’s funeral was held in Rome. The casket, carried on the shoulders of close friends, was accompanied by the sound of Il nostro concerto as it exited.

The designer’s stores, closed in his honor, displayed in their windows:

Issue #30 of Computing Education Things was written while listening to:

🌐 Cool things from around the internet

🔗 repere — exploring data without leaving the browser.

🔗 Flow — bring your unique voice to your writing.

🔗 TurboScribe — save hours transcribing audio with AI.

🔗 Maintenance: Of Everything, Part One — it looks amazing—just like everything Stripe puts out. They’re quickly becoming one of the best publishers when it comes to design.

🔗 Basheer Tome — this portfolio is wonderful, especially the interactive elements. Go ahead and give it a try.

🔗 Quick Links

🎧 Listen to Computing Education Things Podcast

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

Buy Me A Coffee

#29 — Staying relevant without getting overwhelmed

Daniel Prol — Fri, 23 Jan 2026 14:17:56 GMT

Learning CS with purpose

A video featuring four Spanish-speaking software engineers sharing how they’re navigating the AI wave in their day-to-day work. It offers a highly relevant perspective—both technically (as they are software engineers, architects, and developers) and from a business standpoint, with real-world use cases discussed. The conversation provides practical insights straight from the field:

The nature of the shift

If something can be computed, it will eventually fall into AI’s domain. The ceiling isn’t creativity or intent—it’s computability. This reality underscores a broader truth: programming has never been static. From assembly to high-level languages, from frameworks to natural language interfaces, constant evolution has always been the norm. What’s shifting now is not the existence of change, but its pace and surface area.

Yet the magic was never in the code itself—it’s in the business rules, the understanding of real problems, and the solutions we design. Code is an implementation detail. As tools evolve to handle more of that detail automatically, the software engineer’s value increasingly lies in clarity of thought, communication, and product insight.

What this means for Software Engineers

New market niches will emerge. Some jobs will disappear, but new roles and opportunities will open—often in areas we can’t fully anticipate yet. Software engineers are uniquely well positioned to lead this transition. It won’t be driven by veterinarians or lawyers; software engineers sit at the center of the shift.

Fundamentals still matter—especially when it comes to maintaining, debugging, and evolving software systems. AI can generate code quickly, but understanding why something breaks, how to fix it, and how to extend it still requires strong foundations. Senior engineers, equipped with both experience and AI tools, have become exponentially faster. Meanwhile, juniors won’t disappear—they’ll mutate. Knowledge of software engineering principles is becoming more important than mastering any specific language or framework.

Hussein Nasser also studies software engineering by focusing on fundamentals rather than trendy technologies. He starts by asking questions and digging deep. He gives an example involving networks, which illustrates his way of learning—starting from a specific doubt and exploring key related concepts. That’s when deep learning truly happens: when you investigate out of real need.

Hussein Nasser

How I learn Software Engineering

Software engineering is a big and constantly evolving domain with new innovations. While this is true, most -if not all- technologies in software engineering tend to eventually coalesce into few first principals. The engineer is better served learning the fundamentals of software engineering instead of frameworks, languages or platforms that keep changi…

5 months ago · 40 likes · Hussein Nasser

Interfaces, education, and learning

Despite predictions to the contrary, frontend development is not dead. Interfaces may evolve, but user habits, context, and human behavior still demand thoughtfully designed interaction layers. Similarly, how we teach programming matters more than ever. Education must balance foundational knowledge with modern tools—and crucially, preserve the value of struggle. Learning happens by making mistakes, debugging, and understanding why things fail. Removing that struggle entirely risks producing developers who can prompt but not problem-solve.

Being able to express what you’re thinking is now a critical technical skill. Communication—both with humans and with AI—has become central to effective software engineering. AI can either amplify you or shoot you in the foot, and the outcome depends entirely on the user’s understanding and intent.

Focus on what lasts

To stay current: study with purpose. Focus on what you need, ignore most hype, follow signals that last, read foundational papers, and be selective. Remember that LLMs are not all of AI—they’re powerful, but just one subset of a much broader field. Learn the basics well enough to guide AI tools intelligently. Your brain is the best resource you’ll have for the rest of your life—keeping it sharp is not optional anymore. Otherwise, you’re not engineering; you’re just reacting.

The best AI learning resources in 2026 span foundational books, research blogs, university courses, newsletters, and papers. Together, these resources provide both theoretical foundations and practical insights into modern AI systems.

ByteByteGo Newsletter

EP198: Best Resources to Learn AI in 2026

Real-Time AI at Scale Masterclass: Virtual Masterclass (Sponsored…

4 months ago · 183 likes · 2 comments · ByteByteGo

An optimistic outlook

Soft skills are now non-optional. Communication, context-setting, and product thinking are fully part of the job. But this shift also brings opportunity: less meaningless labor, more leverage, more impact—and we’re only at the beginning.

The role of the software engineer is evolving—from assembly, to high-level languages, to natural language, and now toward orchestration. We’re moving from writing everything by hand to directing extremely powerful tools. The question is not whether this transition will happen, but how thoughtfully we navigate it.

Students and AI Literacy with Annette Vee

How do students really feel about generative AI and learning? What kind of guidance are they looking for from their instructors? And how can educators better understand students so they can adapt together to a world shaped by generative AI?

This week on the Intentional Teaching podcast, Annette Vee, Associate Professor of English at the University of Pittsburgh, joins Derek Bruff for a wide-ranging conversation about students, AI, and teaching.

Annette and her colleagues have spoken with hundreds of students at Pitt about AI, and she shares what she’s learned about how conflicted many students feel about AI’s role in higher education—and what it takes to have open, productive conversations about a topic that can be surprisingly difficult to discuss.

The conversation is part of an ongoing collaboration: Annette Vee and Derek Bruff are co-authors, along with Marc Watkins, on the forthcoming The Norton Guide to AI-Aware Teaching.

Like Marc, Annette was experimenting with LLMs before most educators had heard of ChatGPT. Her research sits at the intersection of computation and writing, and her first book, Coding Literacy: How Computer Programming Is Changing Writing (MIT Press, 2017), reflects that interdisciplinary background. Annette was directing the composition program at Pitt when ChatGPT launched in late 2022, and since then she has been helping colleagues at Pitt and beyond practice what she calls AI-aware teaching.

The episode covers a wide range of topics: how computational literacy is changing in light of AI, whether “AI literacy” is even the right term, what Annette has learned from listening closely to students over the past few years, and why AI needs a place in the college curriculum. Along the way, there’s also a brief detour into early programming experiences with BASIC and Logo in the 1980s.

You can listen to the conversation here:

🔍 Resources for Learning CS

→ MIT 6.S191: Introduction to Deep Learning

This fully open course is perfect for anyone looking to either dive into or refresh their deep learning knowledge.

→ The Concise TypeScript Book

An open-source book that offers a clear guide to TypeScript, suitable for both beginners and experienced devs.

→ Compilers and Interpreters Recourses

Here are some great options, organized by difficulty level:

Easy: Crafting Interpreters — Free and beginner-friendly
Medium:
- Writing An Interpreter in Go — Paid
- Writing A Compiler in Go — Paid
Advanced: Let’s Build a Compiler by Nora Sandler — Paid

→ Refactoring English

Refactoring English by Michael Lynch argues that developers aren’t bad writers—they just haven’t practiced deliberately. It shows how to apply programming-style best practices to write clearer docs, blogs, emails, and tutorials that developers actually read.

→ Data Science Resources

A website curated by Nicola Rennie that brings together free data science resources with filters by resource type and categories—making it easy to quickly discover useful and up-to-date materials.

🔍 Resources for Teaching CS

→ Computer Science II Lectures

Chris Bourke from the University of Nebraska–Lincoln is uploading his Computer Science II lectures to his YouTube channel. This is the first one. He previously uploaded these lectures during the Spring 2024 semester.

→ University of Virginia Teaching Hub

The University of Virginia Teaching Hub has recently published several new curated collections developed by educators from UVA and partner institutions. These resources cover a range of timely teaching topics, including strategies for supporting collaborative learning and group work, approaches to dialogue and deliberation across differences, concrete practices for fostering civil discourse—particularly during the first weeks of class—and innovative grading and rubric strategies for assessing student–AI collaboration. Together, these collections offer practical, research-informed guidance for instructors navigating contemporary classroom challenges.

→ Teaching Real-World ML Systems at Scale

This post is packed with concrete lessons for teaching real-world systems, data pipelines, and applied ML.

→ FossFLOW lets you draw isometric diagrams

If Excalidraw (mentioned below) feels a bit flat, this open-source editor allows you to create 3D graphics.

→ An open Algorithms Book

The Algorithm Codex is a free, open-source, literate-programming–style reference that explains undergraduate algorithms with intuition, proofs, and clear complexity reasoning—designed to complement textbooks.

→ 15.773 Hands-On Deep Learning

A graduate-level computer science course designed to equip CS faculty with up-to-date (well, up to two years ago) material on deep learning.

→ The Value of Agentic AI in the Research Lifecycle

These findings will serve as the foundation for a 2026 phase focused on the collaborative development of an open, technology-agnostic blueprint for high-priority agentic AI tools.

🦄 Quick bytes from the industry

→ Rui Fu’s Career Journey

Rui Fu (aka @raycfu) transitioned from a pre-med and biomedical research background into software engineering, advancing quickly through roles at Instagram and Meta. He now focuses on building in public while working on a stealth product. This conversation highlights four key ideas especially relevant for today’s computer science students and early-career software engineers:

Optionality over passion: Rui’s pivot into computer science wasn’t driven by a lifelong passion but by the desire for greater optionality. Coming from a pre-med and biomedical research background—and having struggled in AP Computer Science—his first real signal came when he landed a Nordstrom software engineering internship. There, he discovered that transferable building principles could scale across roles, eventually leading to Instagram and Meta.

Signaling over raw ability in recruiting: In Rui’s experience, landing a role wasn’t just about solving problems—it was about making reasoning visible in interviews. After a rejection from Microsoft, he realized the importance of articulating his thought process. He leaned heavily on referrals to escape the cold-apply pile and later used building in public to create visibility beyond résumés.

Judgment as the differentiator in the AI era: Rui views LeetCode as a screening proxy rather than a reflection of real-world software engineering. He emphasizes system design as the core on-the-job skill and recommends using tools like Excalidraw to sketch architectures and reason end-to-end. As AI-assisted interviews emerge (including at Meta), he argues that the most successful software engineers will be those who can question, audit, and reason about AI outputs rather than follow them blindly.

A nuanced take on the CS degree: Rui still encourages students to pursue a computer science degree, given that software increasingly underpins every field. However, he cautions that the degree alone is no longer enough—what matters is using it to get the principles down, especially as AI automates more routine coding work.

→ From India to Oracle

In this conversation with Amin Manazir, Shreyas Kulkarni explains why he left a comfortable life in India—less for “escape” and more for curiosity, opportunity, and a chance to experience a different culture—especially in a system he describes as intensely competitive and status-gated (where big tech recruiting often concentrates on IITs and top schools).

After studying CS at SRM IST Chennai and working at a fintech company, COVID delayed his plans, but he arrived in the U.S. and deliberately chose San José State University for one reason: ROI. By landing a TA role, he cut tuition from roughly $8.5k/semester to about $1.7k, earned a monthly stipend, and says he finished the master’s net positive (about $12k total tuition, plus earnings).

Being in Silicon Valley helped: SJSU’s practical curriculum and local recruiting pipeline led to multiple internship offers, including Oracle (with a signing stipend and strong hourly pay). He didn’t get a return offer from his intern team, but leveraged the internal recommendation to interview again and land a full-time Oracle role—later growing compensation substantially (he cites ~$300–$400k, much of it in stock).

Along the way, he shares a blunt playbook for breaking in—optimize for a parsable, keyword-smart resume, use referrals as a numbers game, practice LeetCode strategically, and “spoon-feed” interviewers clear evidence of ownership—while also acknowledging the tradeoffs: the U.S. can be high-upside but stressful, shaped by hustle culture, layoffs, and immigration uncertainty.

Notes I’m still thinking about:

Public vs. Private Universities (this section ties in really well with UH and my experience)

You don’t end up making 10× more money just because you paid 10× more for your degree.
In big tech, people from MIT, Stanford, public universities—all end up on the same team. The market doesn’t care how expensive your tuition was.
If your goal is employment, not prestige, public universities can be the smartest financial decision you make.
Brand names help you get interviews, not outcomes. What you do once you’re in matters far more.

AI-Assisted Interviews & Judgment

AI-assisted interviews aren’t testing if you can use AI—they’re testing if you know when not to trust it.
Companies aren’t looking for prompt engineers. They’re looking for software engineers who can reason when the model is wrong.
These interviews simulate real work: messy codebases, partial answers, and tools that don’t always help.

Interviewing even when you’re employed

Interviewing today is like training as an athlete—you don’t stop just because you won the last race.
The market is volatile enough that staying interview-ready is part of the job now.
Most people who get laid off didn’t do anything wrong. The only real control you have is preparation.

🌎 Computing Education Community

There are some really interesting tech talks happening here. If you live in Madrid or you’re passing through, don’t miss it!
Dwarkesh Patel is looking for remote scouts ($100/hour) to help find exceptional guests for his podcast.
The AI Council 2026 speaker application deadline has been extended to January 30.
CS faculty are invited to submit undergraduate course materials that thoughtfully integrate generative AI, with selected submissions featured at an invitation-only GenAI in CS Education Workshop in San Diego (March 16–17, 2026).
EduPar’26 invites papers, posters, and classroom-ready assignments on teaching parallel and distributed computing—including work integrating generative AI—co-located with IPDPS 2026 in New Orleans, with submissions due February 1, 2026.
This group invites faculty to join a project exploring how to intentionally develop professional dispositions in computing students through regular coursework, alongside knowledge and skills.
The Raspberry Pi Foundation is hosting a seminar on January 27, 2026, featuring Salomey Afua Addo on teaching neural networks and AI ethics in junior high schools in Ghana through unplugged activities and storytelling.
FIE 2026 invites abstracts and session proposals on engineering and computing education in the large language model era, with abstract submissions due January 26, 2026, for the conference in Paphos, Cyprus.
Anthropic and Teach for All are partnering to bring AI tools and teacher-led, community-driven training to educators across 63 countries, emphasizing global reach, local context, and teachers as co-creators.

🤔 Thought(s) For You to Ponder…

We’re living in an incredible time for content—there’s a nonstop stream of it every single day. At the same time, we’re completely flooded by it. For me, the real value lies in filtering, connecting, and making sense of it all within my own context. It also matters what kind of content we consume. We need to be critical, selective, and take an analytical approach. I try to do that every day, though it takes time and effort. I’m also fortunate in that I aim to make everything I consume useful in some way—whether personally or professionally. A beautiful reflection.

What does it mean for literature if machines can convincingly replicate a writer’s voice—or even improve upon it? This is an excellent piece from The New Yorker. It’s refreshing to see such careful attention to detail in an article like this!

+1TB. Over 20 million customers exposed (ID, bank info...). It took 2 hours and 30 minutes to steal the data. Self-taught young people. Fantastic radio report. The story behind the Endesa cyberattack.

Uncle Bob and John Ousterhout go head-to-head to solve—well, nothing. They just keep arguing, as usual. You can check out the follow-up discussion on GitHub or watch it on YouTube—whichever you prefer. BTW, I came across this podcast through the Stack Overflow newsletter, and it’s one of the few times I’ve felt the urge to binge the entire back catalog all at once.

This move by Accenture sends a very clear message: AI is no longer just an add-on to its portfolio — it’s the core of its next phase. Just look at who they appointed as CTO: the former CEO of Faculty, an AI-native company.

Paul Stamatiou’s 2025 Year in Review is a thoughtful, personal reflection on a year of major resets—leaving Limitless ahead of its Meta acquisition, taking time off, and joining Sesame—paired with sharp insights on craft, taste, and working with AI (especially his “year of Claude Code”). Concise, human, and refreshingly honest, it blends career reflection with the small details that actually shape how we live and build.

This isn’t the first time I’ve read that working with Claude Code is shifting people from being individual contributors (ICs) to agent orchestrators. It touches on the tension between efficiency and craftsmanship, pushing us toward a kind of meta-work that’s more like what managers do.

Molochinations

AI Killed the Individual Contributor

A friend and I recently discussed how the qualitative experience of coding has fundamentally shifted in a key way these past few months. Engineers morphed from early-2025 snickering at MBAs vibe-coding themselves simple demos that’d surely never reach production to late-2025 embracing of AI as an integral part of professional software engineering…

5 months ago · 93 likes · 12 comments · Philip Su

Are we suffering from “agentic psychosis”? I think it’s too early to say something like that — plus, addiction is a serious issue and not something to take lightly. But this article by Armin Ronacher really gives you something to think about: coding with AI agents feels productive, but does it actually degrade our judgment and the quality of the software? It also points out that sloppy PRs/issues put a burden on maintainers, not to mention the hidden costs and a culture of endless “loops” without proper quality control. So many questions. Honestly, I’m a bit confused. How about you? Share your thoughts in the comments.

The Bitchat app (created by Jack Dorsey) is allowing Ugandan and Iranian citizens to bypass internet shutdowns imposed by their repressive governments.

According to Affleck, AI ‘has no taste,’ and it’s taste that defines art. I wasn’t too far off in this past issue. I believe that the more omnipresent AI-generated content becomes, the more people will come to value real, human-made creations. And maybe that’s the whole point: to gain productivity and automate what we can, so we can focus on bringing the human perspective — the part that’s rooted in life and lived experience.

Written instructions remain, reduce cognitive load, and allow for repeated reference as needed.

Beyond the Scope

Why are "written instructions" a common suggestion for supporting neurodivergent students?

Note: After I wrote the bulk of this blog I posted on LinkedIn, asking my network “Why do you think that providing written and or visual instructions is a good practice for supporting neurodivergent …

4 months ago · 4 likes · 3 comments · Sarah Silverman

📌 Research Corner

The spring semester started this past Tuesday! I’m excited to dive into the two courses I’m taking on campus—Research Methods and Advanced Numerical Analysis.

At some point this year, when my workload lightens up, I’d like to build a side project similar to Booklist—a curated collection of the most recommended reads online, but focused on tech books. Until then, you can check out the latest updates on the list in my Google Docs.

In order to improve my English skills (my goal is to speak like a native), I subscribed this week to this app built by Xiaofeng. It provides curated news articles tailored to my English level. Each article includes a cover image, vocabulary explanations, and interactive quizzes to help learners improve their English reading skills. I’m really enjoying the experience so far—let’s see if I stick with it!

Playing with Gemini 3 and Nano Banana to refine my TikZ graphics for the EDM paper this week—it’s an insane combo. For Cursor users, here are some tips on how to use the tool, shared by the CEO.

Teaching gets better when it’s done collaboratively, and watching others teach is one of the most powerful forms of professional development, as Mark Guzdial writes. I’m excited to start TA-ing Software Design next week under the supervision of my advisor, Amin! I can’t think of a more relevant course for the times we’re living in—where judgment, taste, responsibility, and domain knowledge remain irreplaceable.

🪁 Leisure Line

After having dinner there last Saturday, Vieng Thai is now my favorite Thai food I’ve had in Houston. Highly recommended—no affiliate link. We played Exploding Kittens afterward at home. Big fan.

📖📺🍿 Currently Reading, Watching, Listening

ORION is a short film about Orion Miller, a surfer in Montréal who braves frigid winter temperatures to ride the endless waves of the St. Lawrence River. It’s beautifully shot, and one surreal moment shows Orion floating on a large chunk of ice to get into position for the perfect wave.

Last Sunday, January 18, marked the beginning of the Week of Prayer for Christian Unity, which concludes this Sunday, January 25, with the Feast of the Conversion of Saint Paul. A divided Church cannot fully and authentically bear witness to the mission of Jesus. The lack of unity weakens everything. Unity isn’t the same as uniformity, but Jesus repeatedly emphasized how essential it is for us to be united in what truly matters. Take this example: Jesus chose twelve apostles who were very different from one another, yet He called them to unity — to row in the same direction when it came to what was important. It’s well worth remembering something Pope Benedict XVI once said: unity requires conversion. Unity doesn’t mean asking others to make up for what they lack — it means I need to contribute what I myself am missing: more humility, a greater willingness to listen, less of a need to impose my views or have the final word… so many things that require humility and the inner conversion Pope Benedict spoke of, so that true unity can shine through.

A necessary story. What a powerful episode. The risk involved in telling the truth about the Colombian armed conflict. Go here for the English version!

Seeking Beauty is a new documentary series now streaming exclusively on EWTN+. Hosted by actor and filmmaker David Henrie. Episode 1 focuses on Vatican City, and Episode 2 journeys through Rome, highlighting masterpieces, local voices, and how beauty points toward God.

🌐 Cool things from around the internet

🔗 Chromatic — A colorful daily puzzle game, created by Kevin Feyder.

🔗 flagstories.co — A striking visual breakdown of flag designs, created by ferdio.

🔗 Chosic — A free music discovery and playlist tool hub that helps you find songs, analyze your Spotify stats, generate names, and explore genres in fun and creative ways.

🔗 maptoposter — Transform your favorite cities into beautiful, minimalist designs.

Quick Links 🔗

🎧 Listen to Computing Education Things Podcast

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

Buy Me A Coffee

#28 — Paying attention to see the world more clearly

Daniel Prol — Fri, 16 Jan 2026 11:33:25 GMT

There are issues that need constant attention

"The Philistines Place the Ark of the Covenant in the Temple of Dagon," Battista Franco, Italian, ca. 1540. A moment of collective attention—something increasingly rare in our classrooms.

For my next project, I’m exploring different approaches to addressing the problem of student engagement in the classroom. I’m particularly interested in technologies that can assign in-class tasks to help keep students actively involved.

A recent podcast between two education leaders in Latin America—Freddy Vega (CEO of Platzi) and David Escobar (Director of Confama)—really helped me reflect on this topic. I’d like to highlight a few thought-provoking ideas from their conversation:

On generosity and quality time (recognizing our vulnerability):

So far, the only thing that’s worked for me is physically separating myself from the device. If I have a meeting in a room 20 steps away, I leave my phone on the table and take the opportunity to charge it. One day, I realized I had hit rock bottom. There are two moments that made it clear: when you’re in an important family or intimate moment and you reach for your phone—that’s when you know something’s wrong. The other was work-related. I was in my own board meeting, during a key discussion where I needed to focus—I’m the director—and I grabbed my phone. LinkedIn, obviously. I thought, I need to do something. This is urgent.

On designing spaces and contexts that protect students’ attention and help them make better decisions:

It’s not all about willpower, discipline, or self-control. It’s about creating environments that encourage learning. We need to design educational spaces, companies, and social spaces that support this—like having lockers at the entrance of classrooms.

On ruining everyone’s experience (a principle that applies to universities, work, and relationships):

Why can’t you take out your phone during a play or a movie—even though some people do? It’s a destructively desperate act. You’re watching a movie and someone nearby has their phone at maximum brightness.

On the degradation of attention:

There’s an entire generation that can’t watch a full movie [...] that can’t sit through a one-hour meeting—let alone read a book that demands weeks of sustained attention.

On the power of policies and agreements:

I think we’ll need to start creating agreements like we’ve done for issues like alcohol, tobacco, and driving. Children already face many restrictions—not because we want to limit them, but because they’re not adults yet, and their brains are still developing. While governments and legislatures catch up—which always takes time, and even more so with technology—we’re winning small battles, building arguments, and, sorry for the drama, saving lives.

On technology vs. non-technology:

It’s about distinguishing genuine educational tools from digital distractions—like social media, video games, and addictive content.

On the loss of social skills:

What happens to a young person who’s spent their whole life gaming, with little time outdoors, no street play, and no experience resolving conflicts? What happens when they go out into the world, fall in love for the first time, or get their first job? They can’t handle that beautiful, complex thing—another person’s humanity. They’re terrified because they’ve never had a simple neighborhood fight during a soccer game.

On AI’s effects on productivity:

In the knowledge industry, many people are using AI quietly. A task that used to take an hour now takes a minute—but they still use the extra time to scroll.

If you have any project ideas or think we could collaborate on this topic, feel free to drop me a line at dprol@uh.edu. Watch the full conversation here:

The hard part isn't writing code

AI generates code faster than our ability to understand it. Jake Nations from Netflix warns about this in his talk at the AI engineer conference: We’re shipping code to production that we don’t fully understand. But our programming experience isn’t useless at all; it gave us judgment and taste, it shaped our style. AI makes the mechanical parts “easy,” but it can’t distinguish elegant architecture from technical debt.

The problem is that every time we choose the easy path, we’re trading short-term speed for long-term complexity. And AI doesn’t know the difference between your core business logic and the accidental technical debt from past decisions — it replicates everything with the same confidence. Read more here.

This is where people who understand systems become essential. Programming is still about design, and design requires human thinking.

The hard part isn’t writing code—it’s knowing what to build, how to build it, and how to make sure we built it right. That part is still on us. Nations closes his talk with a reminder: the thinking, the synthesis, and the judgment are still ours:

That’s where Ralph comes in.

Ralph is an autonomous AI agent loop that encodes your judgment into a repeatable system. It runs Amp in clean contexts, grounded in your product requirements (PRDs), and commits working code only after passing tests and typechecks. Between each iteration, Ralph stores learnings in progress.txt and updates AGENTS.md, creating a persistent memory that amplifies design decisions over time.

The human touch is still critical here: writing clear PRDs, refining story scope, tuning prompt behavior, and monitoring Ralph’s output...

This week, I’ve been also thinking a lot about this shift that’s becoming impossible to ignore: the real advantage in software engineering no longer lies in writing code, but in understanding it.

The first video looks at the rise of vibe coding. It shows why this approach can be incredibly fast and effective—especially for experienced developers—but also why it’s risky for juniors and beginners. The core problem isn’t AI itself, but over-reliance on it. When developers generate code without truly understanding it, they can develop an illusion of competence that quickly collapses the moment something breaks. Debugging, reasoning about trade-offs, or adapting the system becomes nearly impossible. The video proposes a simple but powerful principle: Vibe, then Verify. Use AI to generate code, but always follow up by reading it, understanding it, and changing it.

The second video zooms out and places this phenomenon in a broader context: the evolving landscape of software engineering in 2026. The speaker reflects on how things have changed since 2022, when early models like GPT-3 appeared, and how by 2025 tools such as Codex, Cloud Code, and models like Opus 4.5 made it possible for almost anyone to build applications in minutes. This democratization has dramatically lowered the barrier to entry—but it has also raised the bar for standing out.

As a result, the role of the software engineer is shifting. Success increasingly depends on planning, orchestration, and system-level understanding rather than on syntax. At the same time, junior engineers face a tougher market, with tighter budgets and higher expectations. Strong fundamentals, continuous learning, and fluency with modern tools are no longer optional. The video also raises concerns about declining code quality and craftsmanship, as AI-generated code tends to be repetitive and shallow. To remain competitive, developers must become more well-rounded—strong in design, product thinking, and communication—while still maintaining a deep understanding of computer science fundamentals.

That said, we can’t be naive either—we have to keep in mind what Cal Newport points out in this post about the risk of digital deskilling. One thing that helps me is taking a moment to pause and look beyond the initial excitement. And we also can’t forget that there are leaders with direct interests pushing different agendas.

🔍 Resources for Learning CS

→ Cybersecurity Resources

If you’re into cybersecurity, Daniel Miessler is the person to follow. He has a YouTube channel, blog, newsletter, podcast, and you’ll find more resources on his website. Check it out if you’re interested in the topic. Neo Kim also wrote this week about cybersecurity fundamentals.

→ How Browsers Work

What happens when you enter an address in your web browser? An interactive guide to get back to basics.

→ SQL puzzles

Master SQL through progressive challenges. Complete each category to unlock the next.

→ Effective AI-Assisted Coding

Eleanor Berger and Isaac Flath share practical workflows for building with AI: specification-first planning, strong context and documentation, smart model/tool choices, cost control, and the emerging role of agent skills—ending with concrete takeaways for teams shipping with AI today. Watch here.

🔍 Resources for Teaching CS

→ Strategies for Keeping AI (Mostly) Out of Your Course

A great post for instructors who want to resist AI in their teaching. It features UVA professors James Wyckoff (Education) and Ethan King (English). They shared their experiences directing students away from unhelpful AI use in their courses, and CTE associate director Derek Bruff added a few thoughts of his own on the subject. The article discusses both why and how you might help your students avoid AI, and it features Derek’s current taxonomy of AI resistance strategies. You can find it on the UVA CTE website.

→ What a CI/CD Pipeline Really Looks Like

If you teach Software Engineering, students often confuse CI, CD, and Continuous Deployment. This hands-on tutorial breaks down the differences and shows you which approach to teach for different scenarios.

→ A Modern Take on Python Tutor

memory_graph is a modern, local-first reimagining of Python Tutor—a teaching tool and debugging aid that helps visualize how Python’s reference model works. It’s especially useful for explaining concepts like mutability, aliasing, and the differences between assignment, shallow copy, and deep copy—since you can actually see shared objects and how changes ripple through them. It also excels at showing recursion (each function call gets its own set of local variables on the call stack) and at visualizing data structures, rendering your objects’ structure in real time as your code runs. Want to get hands-on? The data model exercises are a great place to start. Big thanks to Bas for building this!

→ Readings in Databases - 5th edition

An opinionated take on both classic and cutting-edge research in the field of data management.

→ Teaching Data Visualization with AI

Enrico Bertini shows how Claude can create interactive visualization teaching tools such as explanatory prototypes for dataviz topics, which could enable the creation of an online repository of interactive teaching materials that would transform visualization pedagogy by eliminating the traditional bottleneck of creating custom interactive examples. See also examples from UW-Madison’s CS765 Data Visualization course: Fall 2025 | Fall 2024. This Carnegie Mellon’s data visualization course also incorporates AI tools.

→ A Conversation with Linda Liukas

Linda Liukas joined Samuel Arbesman on Orthogonal Bet podcast to talk about how children, educators, and cities can relate to computing. If you enjoy smart, cross-disciplinary discussions, this podcast is for you.

🦄 Quick bytes from the industry

→ VS Code in 2026

Kai Maetzel, Engineering Manager for Microsoft’s Visual Studio Code team, joins the Software Engineering Daily podcast to unpack how VS Code evolved from a lightweight, open-source editor into a global, AI-powered development platform. The conversation explores the delicate balance between helpful AI and preserving developer flow—from early multi-line completions to today’s agentic coding models. Maetzel shares how the team measures value vs. annoyance with detailed dashboards, why different interaction modes (foreground, background, and cloud agents) demand distinct UX and tooling, and what this means for cost, trust, and productivity. Despite rapid AI advances, one principle remains central: humans stay at the creative and decision-making core of modern development workflows. Listen to the full conversation below:

→ NeurIPS 2025: Insights from Cartesia’s CEO

This is an interesting conversation from NeurIPS. I particularly liked the part where they discussed how research-driven teams—like Cartesia—turn deep theory into practical products, starting with voice AI. Watch now.

→ From GitLab to Kilo Code

Adam Stacoviak and Jerod Santo from Changelog are joined by Sid Sijbrandij, founder of GitLab, who led the all-in-one coding platform all the way to its IPO. In late 2022, Sid was diagnosed with bone cancer, which started a journey he’s been on ever since—a journey he shares with them in great detail. Along the way, Sid has continued founding companies, including Kilo Code, an all-in-one agentic engineering platform, which he also tells them all about.

→ Adrien Friggeri’s Career Journey

Very inspiring. His experiences at Meta and Clubhouse (being the data guy for other stakeholders) reminded me of my past experiences working for startups. I always say that was a time in my life when I grew a ton professionally—I made mistakes, but I learned by doing.

🌎 Computing Education Community

EduCHI 2026 (May 20-22, Toronto, hybrid) calls for submissions. Paper registration deadline: February 8, 2026; submission deadline: February 15, 2026. Pedagogy Workshop (May 20) for doctoral students, postdocs, and early-career faculty in HCI/Design. Application deadline: February 15, 2026.

Snap Research’s User Modeling & Personalization team is hiring a Research Scientist. They work on behavior and language modeling for personalization across ML problems including retrieval, ranking, and safety, with end-to-end research ownership from idea to publication and deployment.

Pacific Lutheran University in Tacoma, WA is hiring a Program Director for their new M.S. in Cybersecurity (launching Fall 2026). The role involves leading the program, designing courses, and recruiting students.

George Mason CS is recruiting tenured/tenure-track faculty across all areas (including Software Engineering and HCI), as well as teaching-track faculty.

Seeking U.S.-based CS1 instructors who use Canvas to complete a short survey and optional interview for a research study led by PhD student Abigail Liu. If you have any questions, please email Abigail Liu (apliu@udel.edu).

Last day to apply to speak at AI Council 2026. If you’ve built AI systems in production and want to share real-world lessons with the community, apply here.

🤔 Thought(s) For You to Ponder…

Aaron Francis (Database School) sent an email last Friday about how his perspective on AI coding tools has dramatically shifted in recent months, particularly after using Gemini and Claude Opus 4.5, leading him to build multiple desktop apps, games, and a SaaS product without opening his IDE in two months. He’s pivoting from traditional technical education (observing Laracasts and Tailwind layoffs) toward teaching engineers how to use AI tools responsibly, launching a new AI-focused platform while documenting his journey building real products for paying customers.

This excellent article (in spanish) really got me thinking: My admired philosopher José Quintano wrote about bringing out the best in each student, recognizing their talents, and helping them feel valued. Do you believe in your students? I’ve translated his words below:

One of the keys to education is believing in people, one by one. Bringing out the best in each person is only possible if there’s someone who believes in them. This should be applied in the family itself—the place par excellence where each person is unique, valid, and unquestionable for that very reason—and in school, where everyone has their own talents, which we must know how to draw upon to help each person grow.

The impact of AI coding isn’t uniform. It impacts markets with elastic demand and skill constraints (like internal tools, personal software, research) but barely affects markets constrained by regulation, politics, or fixed demand (government, safety-critical, enterprise SaaS). Great analysis by Dave Griffith.

Dancing with Robots: A Software Architect's Journey

Eight Software Markets That AI Will Transform Differently

Chris Loy recently published a thoughtful piece called “The Rise of Industrial Software” that asks the right questions about what impact agentic coding is tools will have on the quality and quantity of software produce. He draws on Jevons paradox, an economic construct about just how markets change the cost of production drops, and shows historical para…

5 months ago · 24 likes · Dave Griffith

It’s not a new idea, but it’s worth remembering that even though AI makes it possible to avoid the hard parts of the creative process (”the suck”), that’s precisely what the work is all about. I appreciated this defense of intentionality in creative work.

It hurts that such a widely used and popular frontend library does not have good financial support, when it is used in many large corporate projects. All my support to the Tailwind team in this difficult situation. I hope they recover.

Adria Blancafort shows here that training LLMs is no longer exclusive to big labs. With the right scaling laws and quality open datasets, anyone with computational judgment can build something interesting and demystify these models.

Jorge Carrión is a Spanish cultural critic and writer I really like. This week, he shared his thoughts on AI and writing in this podcast. He mentioned a recently published academic paper that shows many people already believe AI writes better. A major study involving trained readers found that, for many of them, AI-generated texts were judged to be better than literary ones. It’s an interesting point: In order to avoid the impression that AI writes better than humans, we need to maintain high standards of quality. Part of the literary world has grown used to increasingly less demanding texts.

Isn’t it ironic? AI is killing Stack Overflow as a community, but it’s saving it as a company. It lives off knowledge created in the past, even as it erodes the conditions needed for that knowledge to be produced again.

Human connections are more critical than ever in today’s competitive job market.

You literally have to apply to attend this conference 😲

📌 Research Corner

→ See you at SIGCSE TS 2026!

Good news: I’ve been accepted as a student volunteer for the SIGCSE Technical Symposium 2026 in St. Louis! I was initially rejected, but yesterday I received an acceptance email! I’m really excited to attend this conference—so many interesting people from my field will be there. I’ll definitely write a recap post when I’m back!

🪁 Leisure Line

This Indiana team is a force. An Indiana vs. Miami national championship is going to be a battle—full of huge collisions. Can’t wait for next Monday!

We had a great time this past Sunday on the last day of the Christmas liturgical season. We stopped for donuts on the way at Shipley Do-Nuts.

📖📺🍿 Currently Reading, Watching, Listening

I have many Venezuelan friends, and I’ve been following what’s happening in the country very closely. This Spanish-language episode with UNT Dallas political science professor Orlando Pérez gives a good sense of what possibilities for transition this political moment in Venezuela really opens up, what factors it depends on, and what role U.S. intervention might play in its near future. It’s a great analysis. Mariana Zúñiga‘s reporting at the beginning of the episode—with real sounds and based in Caracas—is also very touching and helps understand how Venezuelans are experiencing the first days without Maduro in power.

I’m so happy that Bruno Mars is back. After almost a decade without releasing any solo music, he’s made a great comeback with “I Just Might” (I personally really like it). It’s the first single from his new album, The Romantic, which will be released in February. This “strategic move” reminds me of Bad Bunny releasing DTMF early last year, which has allowed him to compete for the global streaming throne. Has it been an intense week? Take a break with Bruno Mars’s new single. It’ll get you dancing.

For everyone who fell in love with Mumford & Sons’ early albums, this track will take you back to how they were in the beginning. Let me know if you like it as much as I do.

This Matt Maher song speaks to the internal struggle between the lies we tell ourselves and the healing truth of faith.

🌐 Cool things from around the internet

🔗 watchsomethingwonderful.com — To find wonderful YouTube videos outside your algorithm.

🔗 English News in Levels — Practice your English by reading news.

🔗 Terminal UI for AWS — A terminal-based AWS resource viewer and manager.

🔗 A Data-driven Trip Through the USA — I think I’ve mentioned this before, but The Pudding is putting out some of the coolest interactive content around right now. Seriously next level.

💬 Quotable

He has most who needs least. Don’t create needs for yourself.

― St. Josemaría Escrivá

Quick Links 🔗

🎧 Listen to Computing Education Things Podcast

📖 Read my article on Vibe Coding Among CS Students

💌 Let’s talk: I’m on LinkedIn or Instagram

📖 Tech books I recommend

As always, if you enjoy Computing Education Things, please like, comment, or share this post! You can also support this work through Buy Me A Coffee.

Buy Me A Coffee

#27 — Fostering human agency in Computing Education

Daniel Prol — Fri, 09 Jan 2026 11:31:01 GMT

Writing code by hand as a barrier against AI dependency

Writing code by hand (pen and paper) isn’t such a useless exercise, folks. It’s a natural barrier against AI dependency. An English professor I know already has his students write their exams and assignments by hand in class. This return to “the old ways” isn’t pedagogical romanticism—it’s a conscious response to a profound shift in learning dynamics.

Most professors I know aren’t trying to demonize AI but rather preserve what gets lost when students outsource their thinking, writing, and creativity. Many professors are adopting strategies like more in-person work, shared pair programming in class, or direct supervision of the creative process. Some admit they feel “on the defensive,” facing the challenge of detecting when a text was AI-generated using detection tools (which are also imperfect).

The AI Oral Exam Debate

An interesting discussion unfolded in the SIGCSE-MEMBERS community when an instructor shared their experience using AI voice agents to conduct oral exams at scale in an AI/ML Product Management course. Students disliked it, but the instructor considered it a successful assessment tool.

The case for AI oral exams centers on pragmatic/logistic benefits: eliminating examiner bias, enabling scalability, and offering adaptive questioning without the inconsistency of human graders. Geoffrey Challen (UIUC) frames this as replacing typed exams rather than human conversations, and plans to pilot the approach. Brian Harrington (University of Toronto) adds that verbal responses make it harder for students to pause and consult LLMs, revealing who actually did the work.

But the critics see something more troubling at stake. Fred Martin (UTSA) asks a foundational question: “Teaching and learning is relational—do we really want to be offloading conversations with our students to machines?” Nicholas Weaver (Berkeley) cuts to the technical problem: evaluation requires checking correctness, not plausibility.

Jérémie Lumbroso (Penn) goes deeper, pointing out that the “Fighting Fire with Fire” framing reveals an adversarial mindset incompatible with education. The original post, he notes, showed zero curiosity about students as individuals—treating them as “cargo rather than navigators.” This is Freire’s banking model of education, now using AI to double down on surveillance rather than facilitate genuine dialogue.

The debate surfaces a fundamental question: Should we use AI to scale assessment mechanics, or does this moment demand we reexamine what assessment is for?

Computing Education in 2026: Insights from Raspberry Pi Foundation’s Hello World Podcast

The latest Hello World podcast episode from the Raspberry Pi Foundation features host James Robinson discussing the state of computing education as we transition to 2026. Joining him are three colleagues from the Foundation—Bobby Whyte (Research Scientist), Laura James (Learning Manager), and Rehana Al-Soltane (Learning Manager)—along with international partners from Kenya, South Africa, and Greece. The episode reflects on 2025’s major developments in computing education, particularly highlighting the rise of data science and the growing importance of digital literacy. International colleagues share firsthand accounts of the barriers they face and the progress being made in their respective regions. Looking ahead to 2026, the discussion offers practical advice for educators and explores bold predictions about how AI, curriculum changes, and emerging technologies will shape classrooms this year. You can check out the episode here:

Learning should be a bit challenging. It should stretch your brain and your capacity. So if it’s not challenging and there’s this option where you can take a shortcut, you’re gonna take it. So how do you incentivize and direct people more towards this ‘it’s worth me struggling because it’s going to be valuable in the long run’? How do we incentivize that more invested thinking in their future?

🔍 Resources for Learning CS

→ CodeMender (Google DeepMind)

DeepMind’s CodeMender is an agent that detects and fixes vulnerabilities autonomously before they can be exploited. It uses Gemini to identify flaws, generate patches, and automatically validate solutions before final human review. It can handle large codebases (4M+ lines of code) and it uses static analysis, dynamic analysis, differential testing, and other methods to find issues. Then a debugger and source code browser generate and validate fixes.

→ Gist of Go: Concurrency is out!

Learn Go concurrency from the ground up with 50 auto-tested exercises and tons of interactive examples. It's a full course + book in one.

→ Side Project Ideas Collection

There are quite a few interesting side project ideas here.

→ HN Software Engineering Blogs

Good examples on this HN post of high-quality software engineering blogs with real-world depth.

→ Career Story + Compensation Resource

Sometimes the best manager is the one who lets you go when they recognize you've outgrown what the organization can offer you. It's not a failure of the manager—it's a limitation of the system. Great career story. He also shared this excellent resource for comparing titles and compensation across big tech.

→ R Tutorials

Learn EDA in R with Mine Çetinkaya-Rundel using Positron IDE. The tutorial uses a real project analyzing how homework deadlines affect student performance and stress. You'll learn to clean, filter, and visualize data with ggplot2 while discovering Positron's workflow features. Great for mastering R data visualization. Andrew Gard, a professor of mathematics and computer science at Lake Forest College, also has tons of free videos with useful R tips.

→ AI-Integrated Programming Assignments

A Rock, paper, and scissors exercise for an intro programming class that integrates AI into the process without removing human agency. This other one using Quick, Draw! is also interesting for a machine learning and neural networks class. Excited to see more CS assignments in the coming months from The AI Pedagogy Project by metaLAB (at) Harvard.

→ Making Software (Shaders Chapter) + Nanda’s Interactive Blog Posts

I already mentioned it in a previous edition, but over the break, I managed to catch up again with Dan Hollick's Making Software book and read the chapter on shaders. This one is really beautiful. Fantastic stuff. Since I'm talking about past editions, I also shared one of Nanda Syahrasyad's interactive blog posts in another edition, but it has more on computer science and web development. You can read them all here.

→ Machine Learning Without Equations

Learn how machine learning actually works without drowning in equations. This book uses clear visuals to explain prediction, distance, evaluation, trees, and preprocessing, plus includes a complete end-to-end project. Free to read online.

🔍 Resources for Teaching CS

→ Netflix’s Responsible AI Guide

I'd encourage you to check out this guide Netflix published on their responsible use of content. I think it's a pretty sensible, responsible, and ethical approach to using AI for content production within their platform and ecosystem. The application to teaching seems clear to me for establishing when it's responsible and when it's not. That said, it's important to provide concrete examples.

→ Cloudflare Database Incident

Real-world example of how a small change (database permissions) can trigger a massive failure. I’m thinking about CS classes where this example would fit: SE (error handling/fail-safes/circuit breakers/incident response), Distributed Systems (propagation/cascading failures/eventual consistency), Databases (ClickHouse/permissions/metadata), DevOps or SRE (monitoring/alerts/rollback mechanisms). I really liked Cloudflare’s transparency in sharing technical details about a problem that had real-world scale impact.

→ Every Version of Windows Explained

If you teach operating systems, this video is worth watching—but even if you don't, it's worth it just for the screenshots alone.

→ Mermaid to Animation Tool (Fanfa)

Turn your mermaid diagrams into animations | There are many parts of CS that can always be turned into a graph. Graphs are easier to understand, easier to digest, and help the professor focus the message when there are many technical details. I'm thinking of a DSA class, for example. This tool converts Mermaid code into animations with colors and dynamic arrows. Very nice, though somewhat limited in its free version. Little tip: You can ask AI to help with Mermaid syntax, then paste it into Fanfa for a makeover.

→ MIT Press Algorithm Books

These books on Algorithms (optimization, decision making, and validation) from the MIT Press look quite interesting to read. Excellent for diving into algorithm theory and core ML algorithms. All of these are available as PDF downloads!

→ Stanford CS230 Deep Learning (Fall 2025)

Stanford has released its Fall 2025 CS230 Deep Learning course for free on YouTube, taught by Andrew Ng and Kian Katanforoosh. Here's the syllabus.

→ Flatlogic Open-Source Templates

Top quality premium templates for free | If you teach web development, Flatlogic just open-sourced 29 templates for React, Angular, Vue, React Native, and Bootstrap. Bold and generous move in the web development templates industry.

→ Intro CS Resources (Subgoals + AI/ML)

If you're teaching intro to CS this Spring, I suggest the subgoals exercises (Python or Java) in Runestone Academy or Learn CS Online. If you'd also like to give students some understanding of how AI systems work, what they can and can't do, and present some of the basic ideas of machine learning, large language models, and so on, check out my podcast episode with Danny Yellin, who's teaching a course on LLMs for software engineering this semester. For Neural Networks, check out this playlist from 3Blue1Brown, and Leland Beck from San Diego State University recommends using Alice to teach programming concepts for non-majors.

→ Software Architecture Patterns Playbook

If you teach Software Architecture or a related course, check out this free Architecture Patterns Playbook and download it if you haven't yet.

🦄 Quick bytes from the industry

→ Rong Yan's Career Journey

Rong Yan has an interesting career path. He did CS in China (Tsinghua) and a PhD in CS at CMU. After a few years in an industry research lab at IBM, he decided to pivot toward pure tech industry in various positions at Meta, Square, Snapchat, Verishop, HubSpot, and now HeyGen. This conversation is very focused on management, but what I take away are two ideas:

His decision to choose Meta (Facebook back then) instead of a CS faculty job or quant firms

After spending over 8 years in research (5 years PhD at CMU + 3 years at IBM Research), Rong faced a critical inflection point in 2009. He had three clear paths ahead:

Academia - CS Faculty position at a top school (he had interviews lined up)
Quantitative Finance - Offers from top quant trading firms
Software Engineering - Join a fast-growing tech company like Facebook

His decision-making framework was:

Impact at scale: As he puts it: “I want to go to places that can make engineering to be the first class citizens. I think in the financial world, engineering is always a second class... I can be a faculty, but being a faculty, I think the impact is smaller because you can only impact the scale of a school or maybe the community, but not the entire world.”

Engineering as first-class citizens: He sought environments where engineering was valued as a primary driver, not a support function. Facebook represented this perfectly - a place where engineers shaped product and company direction.

How he manages, even being at the highest management level, to get into the details of technical problems

This is perhaps the most distinctive aspect of Rong’s leadership philosophy, deeply influenced by Facebook’s culture:

The Facebook bootcamp revelation: When Rong joined Facebook, he experienced their 6-week bootcamp where everyone - including VP-level hires - sat alongside new grads finding bugs, fixing bugs, and writing pull requests. He recalls: “There was a VP-level hire sitting right next to me. And she’s just doing the same thing as what I was doing... She was doing that for six weeks. That actually shocked me because I come from IBM. IBM’s VP never coded anymore.” This experience was “deeply planted in my heart” and shaped his entire approach to technical leadership.

Current technical practice: As a CTO, Rong maintains remarkable technical involvement:

Writes 2-3 pull requests every week
Reviews code regularly
Still writes actual production code

His rationale is clear: “I’m not really good at speaking up if I don’t know the details. I want to make sure that I understand the details so that I know I’m not making things up so that I can make the best strategic decisions for the teams.”

The prioritization framework: How does he balance this with executive responsibilities? His answer is elegantly simple:

“At the beginning of every single week, I will ask myself, what are the top three things I need to achieve? Only focus on those top three things, and then everything else is less important.”

Some weeks, one of those top three things is development and getting into details. The key insight: “I’m not saying that every single week you need to do the same things. But every single week, you need to have a theme for your work.”

First principles thinking: He constantly recalibrates against his North Star goal, understanding how each weekly theme delivers maximum impact. As he emphasizes: “I think the best people, not really just spending more time, but they’re really good at allocating their time. And by understanding the priority of each direction, and spending the right time at each priority.”

Philosophy on technical depth: Rong believes “everyone who works on engineering needs to be technical” - this is non-negotiable. Being technical and detail-driven is critical for making sound strategic decisions. He wants to ensure he’s not “making things up” but rather making informed decisions grounded in deep understanding of the systems and problems his teams face.

→ Steve Yegge’s take on vibe coding

Lots of interesting insights here:

Next year the tools are going to get much better at decomposing the task and assigning them to the right size model for cost optimization.

Just because you don’t have to write code anymore you still have to learn a massive amount of stuff to be an effective software engineer in the new world.

→ Bret Taylor's Advice for the Next Generation of CS Students

Why do I think a CS degree is stronger than ever with AI? A CS degree makes a lot of sense because it provides you with a formal and solid foundation that’s key in the job market, especially with AI. Those fundamentals are what help you make decisions based on essential properties that don’t change as much as tools do. What matters now isn’t so much syntax, but how we ask for things or how we interact with the LLM. In those questions or in the context we give the LLM lies the difference (conditions, technical requirements, etc.).

And not just in the input but in the output—that is, how to interpret the response. AI doesn’t replace experience or human judgment; it’s a great ally ONLY when used on top of a solid knowledge base. It’s easy to become dependent or to overlook subtle errors that are hard to detect. Therefore, what you learn in your degree will help you leverage AI to review code, automate tests, write documentation, and build pipelines with a clear understanding of the environment so you don’t settle for partial or fragile solutions. With this combination of solid foundation plus real experience, you’ll be able to benefit much more from AI.

Here is what Bret Taylor has to say on the subject:

I still think it’s extremely valuable to study computer science. I say that because computer science is more than coding. If you understand things like big O notation or complexity theory or study algorithms and why a randomized algorithm works and why two algorithms with the same big O complexity—one can in practice perform better than others—and why a cache miss matters and just all these little details, there’s a lot more to computer science than writing the code.

I believe that the act of creating software is going to transform from typing into a terminal or typing into Visual Studio Code to operating a code-generating machine. I think that is the future of creating software, but operating a code-generating machine requires systems thinking. And computer science—there are other disciplines as well—but computer science is a wonderful major to learn systems thinking.

And at the end of the day, AI will facilitate creating this software. We may do a lot more in the next few years than we can’t even imagine, but your job as the operator of that code-generating machine is to make a product or to solve a problem, and you really need to have great systems thinking. You’re going to be managing this machine that’s doing a lot of the tedious work of making the button or connecting to the network.

But as you’re thinking about the intersection of technology and a business problem, you’re trying to affect a system that will solve that problem at scale for your customers. And that systems thinking is always the hardest part of creating products.

I think that whether AI is writing code or doing the design or doing all these other things, you need to learn how to have a system in your head. You need to understand the basics of what’s hard and what’s easy and what’s possible and what’s impossible. And AI can help you do that too, by the way. But I do think that’s a really useful skill.

I think computer science, especially the foundations, will continue to be the foundations of how we build software. And understanding that—when you’re interacting with something that’s smarter than you and producing code that you may not completely understand—how you constrain it and how you get it to produce these outcomes, I think it will require a lot of sophistication, actually.

In this episode, Bret also offered a perspective that balances the challenges educators face today with optimism about where we’re headed. I do my best to summarize it in my own words:

Bret Taylor has a thoughtful perspective on AI in education that balances optimism with awareness of current challenges. He believes we’re in an awkward transitional phase, much like when graphing calculators were first introduced to AP calculus exams. The education system hasn’t yet adapted to a world where students have superintelligence in their pockets, and many traditional evaluation methods are broken by LLMs. Teachers are struggling because technology is moving faster than educational institutions can keep up.

Despite these challenges, Taylor is fundamentally optimistic about AI’s potential as perhaps the most effective educational tool in history. He sees it as a democratizing force that gives every child access to a personalized tutor capable of teaching in whatever style works best for them, whether visual, audio, or reading-based. Students no longer need to be wealthy to afford private tutoring, and those who excel beyond their school’s curriculum can now access advanced material that might not be available otherwise. He envisions AI amplifying agency for motivated learners, giving them the equivalent of the best combination of every teacher they’ve ever had.

Taylor actively encourages his own kids to integrate AI into their learning. His daughter uses ChatGPT to explain Shakespeare, take practice quizzes before tests, and his oldest learned to code by consulting ChatGPT whenever she had questions. He’s intentional about teaching them to use these tools constructively, viewing it as an essential skill for their future.

However, he acknowledges a darker possibility: these same tools can enable students who want to avoid learning. The next few years will likely be bumpy as parents, teachers, and the education system figure out how to navigate this new landscape. Yet Taylor remains confident that just as education adapted to calculators, it will adapt to AI through redesigned homework, testing, and classroom approaches.

On the question of phones for children, Taylor draws a sharp distinction. He doesn’t think mobile phones are great for kids and advocates waiting a long time before giving them smartphones. He sees phones as addictive devices with push notifications that don’t belong in schools. But he views ChatGPT completely differently, comparing it to Google search rather than social media. It’s a utility for learning, not an addictive entertainment platform. In his view, parents wouldn’t typically ask “when should I let my kid use Google search?” because it’s simply a different category of tool. For his own kids, AI access comes through computers at desks rather than pocket devices, maintaining that boundary between learning tools and potentially problematic mobile technology.

🌎 Computing Education Community

Kristin (CS-Ed Podcast) has shared many resources on AI and teaching, but now she finally decided to create one podcast episode about it: some of the variability has gone down, more people seem ready to listen, and the paper this episode discusses seems like an important topic we need to discuss more.

Bucknell University’s CS department is hiring an open-rank professional track faculty member (starting Fall 2026) to teach core undergraduate courses like Data Structures, Software Engineering, and Computing Ethics.

Utrecht University is hiring a PhD candidate in Requirements Engineering and Specification-Driven Development.

Do you know any really amazing undergraduates who would like to spend their summer in Raleigh doing CS and AI education research? Please encourage them to apply to our REU Site: Research Experiences in Innovative Computer Science and Artificial Intelligence Education.

Miranda Wei is recruiting PhD students starting Fall 2026 to work on human-centered security and privacy, with focus on sociotechnical safety, online abuse, and social media studies.

CSEE&T 2026 will be held at the University of Florence, Italy, July 20-22. The Call for Papers is now available.

Augustana College is hiring a Professional Faculty position in Computer Science (starting August 2026) - primarily teaching-focused. Application review begins January 15.

Dan (deblasio@CS.CMU.EDU) is looking for examples of coding assignments for non-CS majors (specifically Biology majors) that are designed with AI coding assistants in mind, as his current Jupyter notebook assignments have become too simple with GenAI tools.

SIGCSE Tutorial 104 on FPGAs/configurable hardware and RISC-V (Wed, Feb 18, 7-10pm) - early registrants receive free UPduino FPGA platforms. Contact Bill Siever (bsiever@gmail.com) with questions.

ITiCSE 2026 (Madrid, July 13-15) is seeking additional PC members for full papers, posters, and tips/techniques. Review periods: Jan 12-Feb 18 for full papers, Mar 18-Apr 8 for posters/tips. Sign up!

The ‘Student Participation in Team-based Software Projects’ SIGCSE Affiliated Event isn’t listed in the main SIGCSE registration—you need to register separately. It’s a half-day workshop (Wed, Feb 18, 1-5 PM, Room 105) for instructors teaching team-based software development. Focus areas include HFOSS projects and courses using real clients, large codebases, and industry tools like GitHub. The format includes short presentations and discussions to share teaching materials and approaches.

Susan Rodger is selling Notable Women in Computing playing card decks at SIGCSE TS 2026. The cards (54 different women, one per card) come in two sizes: regular ($5) and large ($10). They can only be purchased during conference registration (through Feb 7) and must be picked up in person at the exhibit hall (Thursday-Saturday, or Wednesday with Susan directly).

CCSC-Midsouth 2026 will be held April 10-11 at Lipscomb University in Nashville, TN, with a submission deadline of January 12, 2026 for papers, posters, workshops, and tutorials, plus a student programming contest (teams of up to 3) with registration due March 27, 2026—reviewers are also needed and should sign up even if they’ve reviewed before.

🤔 Thought(s) For You to Ponder…

It’s ironic—the parallel between AI and what we’re losing with this new way of life. The dehumanization of not cooking (kitchens are more than just a physical space: they’re spaces for socializing; they’re where we connect, where we care for each other by preparing the food we’ve bought, where we learn and pass down knowledge…). It’s culture, it’s memory…

I found this interview with Rob Riemen in Ethic very thought-provoking:

The greatness of Velázquez’s painting is very different from Trump’s “greatness”; spending a night with any sick person, in the hospital or at home, is a different kind of greatness from the “greatness” of the world’s richest man. Why is Elon Musk’s “greatness” so fascinating right now? Because he’s the richest man in the world. We’re obsessed with that kind of greatness, a false greatness because it has no substance or quality—it’s only quantity. It points to a type of power that is ephemeral, that won’t endure, unlike Bach’s music, for example.
But one can maintain their dignity, act well, not thinking of their own benefit but of the common good. That’s why the Muses are important, as important as language and literature, which allow us to know a different world and encourage us to champion dignity. Today’s utopians are tomorrow’s realists—this is already known. We shouldn’t be impressed or depressed by what happens around us. Acting correctly, each of us in our own sphere, is already a complete triumph.
We already know that screens act like drugs—we know it, but we keep using them... And the worst part: we give that drug to children! We are free, I insist, to change the world, for example, with our use of screens, but we can’t just believe it; we must act accordingly.
What use is love? What exchange value does the woman or man we love have? None. The fundamental things in life are not useful; they have nothing to do with utility. Think about what gives meaning to our life—friendship, for example. The moment it becomes instrumental, it loses its inherent value and becomes a tool.

This New York Times article got me thinking: it’s clear that technology is changing our capacity for reading, reasoning, and concentration. And it also seems clear—though we still need more studies and evidence—that maintaining intact cognitive abilities is on track to become a luxury.

I’ve been catching up on the Spanish podcast Monos estocásticos. It’s where I usually go when I want to form an opinion about AI news. I sometimes find it hard to process so much information. They explain dense topics in simple ways.

Recently they interviewed Tíscar Lara, who recently published this book on AI and education, and she said two things I’d like to share with you:

On agency:

We need to be aware of that basic scaffolding so we have the judgment not to skip the process.

On humanities and technology:

Technology itself also builds culture… the world of the humanities also gives us tools for knowledge and analysis and for pausing to think and doing this critical reflection, which I think is also good for technology development.

Going back to issue #1 over the holidays, I was thinking about the concept of ‘flow state’—how there’s something intrinsically human and satisfying about doing things yourself, even when it involves effort, frustration, and mediocre results.

Facing a challenge, testing our skills, seeking out information, learning new things, solving problems, connecting our actions with the outcome... these are activities that make us feel good, that lead us to what Mihaly Csikszentmihalyi called the ‘flow state.’

If we think only about the outcome, there’s an easier, cheaper, higher-quality external alternative for all of it. Whether it’s a specialist we can turn to, an industrial process that makes everything perfectly uniform and hyperproductive, or an AI that automates the process... it’s really hard for ‘do it yourself’ to make economic sense.

And it’s a strange feeling when what makes human sense doesn’t make economic sense, and what makes economic sense doesn’t make human sense. So where does that leave what makes us human? Is there room for resistance? At what cost?

Because reading isn’t running. Reading is stopping. Reading isn’t about speed—it’s about digestion. About engaging with what you read, chewing on it, thinking as you go. Speed-reading is like running through a museum without looking at the paintings. You can count it as an accomplishment, but what’s the point? Reading is a conversation with someone who’s probably no longer alive but still has something to tell you. And that conversation, like good dinners with friends, can’t be rushed. There’s only one real trick to reading better: read with intention. Choose what you read carefully, give it time, go back if you need to, let ideas settle. It’s not sexy, it’s not fast, it’s not scalable. But it works.

La cabeza de Sussudio

La lectura rápida es un timo

Atenti…

8 months ago · 8 likes · 1 comment · Adrián Sussudio

Opinionated article but very solid and from a well respected person in the field. I especially liked his point about Claude Code.

Technology that wants nothing from you. Single-purpose technology. That doesn’t try to keep you engaged longer or get you to do more things—it just does one thing, and does it well.

I hope and expect that models like the University of Dallas, which is making a strong commitment to classical liberal education, will become more and more common in Spain and Europe. UNAV and UFV are clear signs that something is moving in that direction. I see it in the students—they have a much broader vision. I wish there were more CS majors with liberal arts minors. I think this profile is more necessary than ever.

Great stuff from Marc Watkins on centering education around human agency rather than AI fluency, embracing struggle, and building student awareness. I want to put this into practice for myself in 2026, in a more conscious way.

Human connection remains a powerful competitive advantage. Knowing there’s a human being behind the work fundamentally changes how we experience it. Put another way, behind any creation, we don’t just see the result—we see the effort, the intentionality, a window into someone else’s internal experience.

Copilot adoption is a concern internally at Microsoft. The problem isn’t just technical, but also cultural and educational. I know Microsoft is working on training so that many clients (especially companies with more licenses) better understand all the new features.

This post argues that attention is a choice that shapes our lives, and offers 14 practical habits to be more intentional about what we focus on. It’s been a long process, but I’m glad to know I’ve been cultivating most of these habits over time, and the day feels much more productive. I especially liked point 2: choosing content with substance, that’s thoughtful, inspiring—the kind that leaves a deeper, enduring impression and changes how we think and feel. Anything that can nourish the mind rather than distract it.

The Future of Attention

Healthier habits for living more intentionally in 2026

While on a walk in Laguna Beach…

5 months ago · 13 likes · 2 comments · Gloria Mark

You’re not competing with other startups; you’re competing with the rate of capability improvement in the foundation models. For example, Claude Code now has native LSP support, giving it IDE-level code understanding that makes many third-party code intelligence tools obsolete

Dancing with Robots: A Software Architect's Journey

Claude Code Sees Like A Software Architect

Claude Code shipped native LSP support last week. If you’re not a programming languages and development tooling nerd, that sentence probably means nothing to you. Let me explain why it matters, and why it’s quietly devastating news for a whole category of AI coding startups…

5 months ago · 31 likes · 6 comments · Dave Griffith

Oxide encourages LLM use but prioritizes human responsibility—employees remain accountable for all LLM-generated work. LLMs excel at reading comprehension, research (with source verification), editing polished drafts, and code review. They’re problematic as writers because LLM prose undermines authenticity and trust. For coding, they’re useful for experimental work but production code requires human judgment and careful review.

Interesting! The bottleneck has shifted from writing code to proving it works. But as Zarar Siddiqi says in his highly recommended Substack, the good news is that we already have tools popping up that make this easier.

Andy Pavlo’s annual databases year-in-review is back! PostgreSQL dominated 2025 with major acquisitions (Databricks bought Neon for $1B, Snowflake bought CrunchyData for $250M), every DBMS added Anthropic’s Model Context Protocol support, MongoDB sued FerretDB over API compatibility, five new file formats challenged Parquet’s dominance, and Larry Ellison briefly became the richest person in history thanks to Oracle’s AI datacenter deals.

I completely agree with Mattias: AI has also given me back that productivity I lost in the past with so many specialized domains (the frontend was absurdly complex). The good thing is that if you have experience now, you can go from idea to execution in days, recognize good code from bad code by capitalizing on that experience, and be more productive. That mental space for creativity is what has changed with AI, because we no longer have to be mentally saturated with technical details, but rather focused on building, which is something I enjoy when I do web development. It’s much more fun.

📌 Research Corner

→ Alejandro Piad Profile + Substack Reflection

I really like Alejandro Piad's profile (University of Havana). He embodies exactly the kind of CS educator I aspire to be—someone with a public outreach dimension democratizing knowledge through Substack, a tenured academic side conducting research and teaching, an author's portfolio spanning both popular and technical books, and a community builder (AI Cuba). Looking at the raw statistics on my Substack, this post on “Don't Turn Your Brain Off” stands out as receiving far more attention than the others. Somebody reposted this article on LinkedIn, and it picked up like crazy! In this new year, I’d like to keep going on my weekly post schedule. I want to keep writing about “computing education” explaining why this is important and what we can do about it. I will keep exploring the intersection of learning and technology. For sure, you will see more research, but also more about how to make learning and teaching cs experiences more effective, using or not AI.

→ PhD Exploration + Inspiring CS Academics Podcast

Although my focus is on computing education and my research centers around AI in programming and computing education, I like to explore other fields of computer science so I don’t lose touch with other academic worlds. I want to work on problems I find interesting, and I’m always very broad in what excites me, so I generally like to collaborate. I found this podcast with some very inspiring CS academics. I particularly liked this episode with Michael J. Freedman and this one with Jelani Nelson.

→ ACM Webinar on GenAI Ethics in CS Ed

After watching the ACM webinar on this long collaborative paper between several universities on the Ethical and Societal Impacts of GenAI in Comp Ed, my main takeaway is that this study validates the ESI Framework (Ethical and Societal Impacts Framework) for analyzing GenAI ethical dilemmas in CS Ed from multiple stakeholder perspectives, identifying conflict archetypes (such as automation vs. human augmentation, or cognitive offloading vs. critical partnership) that help navigate ethical decisions and evaluate trade-offs. Here are the slide screenshots in case you want to dive deeper into it. For me, listening to Tony Clear, Janice Mak, and Tingting Zhu was a pleasure. I learned a lot about the whole process. As a first-year PhD student, this helped me get my thoughts more organized when it comes to presenting work and considering all the elements.

→ Brain Activity Study (Nataliya Kosmyna, MIT)

Because of my research area, I read a lot of papers about how AI is impacting universities. I have some context since I’m working for one. But I don’t have a clear picture. I’m learning as I go. The integration of tools like ChatGPT in education is transforming the learning experience. Whereas students used to have to grapple with complex texts and solve difficult problems as an essential part of the learning process, today many of those struggles are replaced by instant answers generated by AI. It’s hard to detect, so it can’t be penalized. This trend, driven by our cultural preference for instant gratification (short videos, quick answers), poses challenges for developing critical and deep-thinking skills, and forces us to rethink how we prepare new generations for a world increasingly mediated by AI. As this paper points out, the cognitive debt is real.

And speaking of cognitive debt, this study by Nataliya Kosmyna from MIT Media Lab reveals that using chatbots like GPT-4o to write essays significantly reduces brain activity compared to working independently or using a search engine. Key findings from the paper:

Up to 55% less neural connectivity compared to the group that wrote without assistance, and between 34 and 48% less for those who only used search engines.
Additionally, AI-assisted students showed lower information retention, a weaker sense of authorship, and worse performance when they later had to write without assistance, demonstrating that relying on AI from the start can undermine deep learning.
The study emphasizes that while AI offers opportunities to enhance research and thinking, its premature and unreflective use can weaken intellectual development, which is why it recommends incorporating AI support only after building a solid foundation of individual knowledge. That’s why I’m concerned about its use at early ages—that’s where I believe the greatest impact is.

→ Derek Bruff Resources

I’ve been following Derek Bruff (UVA) for a few months now, and he produces interesting learning & teaching material. He normally doesn’t speak from a technical perspective, but rather from an educational one. For example, here and here he focuses on some of the pedagogical reasons an educator might want to design a custom AI chatbot. I’ve been thinking about ideas for student engagement in the last few weeks—it is a topic I want to research more.

During Christmas, I also listened to him on the Grading Podcast talking about his history with alternative grading practices, faculty development on grading, AI-aware teaching, and the Alternative Grading Institute! You can listen to their conversation here.

🪁 Leisure Line

I had dinner here with a friend. It was so good and tasty 🤤 In case you want to add it to your list, we went to the Vintage Park location.

It was Aristotle who argued that alongside intellectual work—which elevates the person—every free man should learn a manual trade as a way to develop their different dimensions. Christmas was a perfect time to practice with more hands-on work like…. It opened me up to experiencing reality in a way different from the merely virtual world of screens—more truthful, more real. That contact with the real allowed me to develop in dimensions that our digital, virtual, screen-based world overlooks.

This year I had to open family gifts remotely. It was also nice. I like to collect sweet details throughout the year in my Notebook post-it. These were this year’s Christmas gifts.

Unwrapped these presents from under the tree on the 25th! Absolutely happy with my Williams Racing Team polo and cap – what a fantastic season it’s been for Williams finishing P5 in the constructors’ championship! The UH gear is also super cool – Go Coogs! And oh wow, a Texans tee (does this mean I’m supposed to start following pro football now? That’s a whole other sport to keep up with… noo 😂). Also got some other great stuff: a Car Phone Mount, candies, a new mouse, a desk mat, cufflinks, a streaming LED light, and some Happy Socks. Feeling very grateful! The most random gift of the day – though it wasn’t for me – goes to… a turtle!

📖📺🍿 Currently Reading, Watching, Listening

After listening to a recent episode of El hilo about data centers in Chile, I became interested in learning more about the topic, which led me to ‘Colossus,’ a two-episode special from the Search Engine podcast. Both episodes explain the genesis of these centers in various rural regions of the US, how they’ve become indispensable hubs in what’s known as the ‘AI race,’ and the reaction of residents living in these areas that are now filled with buildings created specifically to power AI tools.

Colossus 1 | Colossus 2

Watched The Thinking Game. It’s about the early days of Google DeepMind and gives a great perspective on how AI has evolved. It already has over 200M views on YouTube. I couldn’t agree more with this tweet.

While fantasy isn’t my thing—I usually prefer essays or other types of novels—I discovered these videos from Brandon Sanderson’s writing course at BYU. I can’t recommend them enough. Not only do you hear Brandon deconstruct the craft of writing, but his genuine passion for story and helping people become writers shines through.

It’s surreal. But I loved this musical short with a pretty peculiar protagonist. Life is learning to coordinate differences without erasing them.

🌐 Cool things from around the internet

🔗 into.md — If you’re trying to pass any website to an LLM, you should 100% be using this tool.

🔗 a cute(-ish) koala — Found in Cassidy Williams Newsletter: Alvaro Montoro live coding some CSS Art. All drawn with HTML and CSS.

🔗 VoiceInk — Great voice dictation tool for Mac.

🔗 Try X in Y minutes — It’s not a bad idea to rethink the “better C” language among this family of languages that Anton proposes here. If you decide to go for it, he has created a product to create an interactive guide for it. See examples here.

As always, if you enjoy Computing Education Things, please like, comment, or share this post! You can also support this work through Buy Me A Coffee.

Buy Me A Coffee

Computing Education Things

#45 — An anthropological approach to AI in Computing Education

Reflections

Technology Adoption as a Design Problem

Consciousness ≠ increasing computational power

🔍 Resources for Learning and Teaching Computing

→ A really good free lecture on system design

→ An Introduction to Software Engineering

→ CS51: Introduction to Computer Science (Pomona College)

→ Training for Teaching-Focused Careers

🦄 Quick bytes from the industry

→ CS gives you a highly versatile skill set as the industry continues to evolve

🌎 Computing Education Community

🤔 Thought(s) For You to Ponder…

📌 Research Corner

🪁 Leisure Line

📖📺🍿 Currently Reading, Watching, Listening

💬 Quotable

🌐 Cool things from around the internet

🔗 Quick Links

#44 — How to keep people engaged for hours

Reflections

How to Make Long-Form Lectures Engaging (Online and In-Person)

The knowledge that will matter most in the future

How AI Has Changed Computing Education

Other Topics

Taste vs. Criterion/Judgement: The Definitive Take

The “Flaky Compiler” Metaphor

What happens if the software survives, but no one remembers why it survives?

🔍 Resources for Learning CS

→ Some thoughts on working with AI models

🔍 Resources for Teaching CS

→ In-class writing instead of a semester-long paper

→ ACM Code of Ethics and Professional Conduct

→ MAP-CS (Mission Aligned Programs for Computer Science)

🦄 Quick bytes from the industry

→ Exploring with agents

→ What jobs are AI jobs?

🌎 Computing Education Community

🤔 Thought(s) For You to Ponder…

📌 Research Corner

🪁 Leisure Line

📖📺🍿 Currently Reading, Watching, Listening

💬 Quotable

🌐 Cool things from around the internet

🔗 Quick Links

#43 — Why The Demand for Systems Thinking Isn’t Going Away

Reflections

Programming Throwdown Is Back with an Episode on Agentic Coding

A Brief History of AI-Assisted Coding

What “Agentic” Actually Means

The Context Window Problem

AGENTS.md and Code Hygiene

What Still Goes Wrong

On the Future of Software Engineering

Teaching Faculty Hiring

What skills do people need to successfully code with AI? CS + writing skills

🔍 Resources for Learning CS

→ If LLMs can generate the code, what’s left for you?

→ The Laws of Software Engineering book is out

→ Allocating on the Stack

🔍 Resources for Teaching CS

→ Python IDEs for teaching

→ Distributed Systems lecture series

🦄 Quick bytes from the industry

→ Great episode with Martin Kleppmann

→ I loved the conversation with Félix Ruiz on the Tengo un Plan podcast

🌎 Computing Education Community

🤔 Thought(s) For You to Ponder…

📌 Research Corner

🪁 Leisure Line

📖📺🍿 Currently Reading, Watching, Listening

💬 Quotable

🌐 Cool things from around the internet

🔗 Quick Links

#42 — Teaching DSA in the Age of AI

Reflections

Mentions during the episode:

The Computing Education Things deep dives relevant to this episode:

Where to find Kevin: