Tech v. Tech
8 min read
Computing educators frequently make reference to “industry”, by which we mean the technology industry. Industry demands that we teach C++. Industry requires that we teach ethics. Industry interviews for proficiency with coding agents. The technology industry has never been a monolith, but as long as I’ve been teaching “industry” has had expectations that are coherent enough for this shorthand to make sense. Our relationship with industry shows up both in what we teach and in the technology we use in our classrooms. But I think we’re about to see a split caused by coding agents, one with important implications for computing education.
As an example of what’s happening relevant to educational technology, consider Canvas
Canvas is remarkably poor software.
I’ve never used it for my courses due to its many restrictions that hamper educational innovation.
Recently I had occasion to use it as part of a faculty workshop.
Even though I had low expectations, and was driving the API using Claude, rather than through the web interface, I was disappointed by its limitations.
For example, I had wanted to create an assignment where students
And I’m clearly not the only faculty member struggling to use Canvas effectively.
Having taken several courses at Illinois recently as a student, I can attest that, while it is possible to create a well-organized course using Canvas
Even if you don’t use Canvas, you might be wondering: Didn’t I hear about Canvas recently? You did. Because the site was hacked, resulting in a multi-day outage in May 2026 that caused chaos across higher education. At Illinois, the attack occurred right before finals, and the university took Canvas offline for two days while it assessed the risk. Instructors couldn’t prepare exams and students couldn’t prepare for them. Apparently this is what we pay millions of dollars for. And Canvas clearly has a captive market. In the wake of the May 2026 outage, I never heard the idea of moving away from Canvas even mentioned at Illinois.
Canvas and other software-as-a-service tools exist for one simple reason: people can’t do better. I maintain my course’s online learning platform—but I’m a computing educator, and even among computing educators that makes me unusual. Most make do with the software the university provides and accept its limitations.
But today’s coding agents are rapidly disrupting the relationship between software developers and software users. This is where the split starts: the same disruption that threatens the educational technology industry enables educators to create innovative educational technologies.
This summer I participated in a faculty workshop on AI run by the University of Illinois’s Center for Innovation in Teaching and Learning (CITL). Watching what faculty can now create by working with coding agents was astonishing and inspiring. One of my colleagues, John Clark, entered the workshop with months of experience using AI to support his teaching. With little technical background he’s been building an entire suite of architectural design tools, including a whole set designed to support a vision-impaired student. I’ve spent hours watching in awe as he demoed all of the things he’s created.
Another colleague, Ben Clegg, arrived never having used AI. Two weeks later he was using Claude to create Socratic tutors for his biology course and deploy them online. He kept asking me: How do I do this? I’d say: Claude will walk you through it. He’d show up the next day with it done. Out of the five faculty I worked with closely, all were able to create remarkable tools using AI coding agents.
Why should faculty continue to use Canvas at all? We made do with an expensive and shoddy product because we couldn’t do better. But now we can. And we should! We’d need support from our university IT to provide access to internal university data (like enrollment) and secure interfaces for handling sensitive data (like grades). But this could be done at a fraction of the cost of what we pay for Canvas. The resulting platform would provide faculty with incredible opportunities to build effective and innovative courses and materials, rather than the limited mediocrity enforced by current LMS platforms. This is already possible, and will just get more and more tractable as coding agents and our ability to use them effectively continue to improve.
This is just one example of a new opportunity available to every professional who uses computing technology. Today, domain experts use tools built by people who know a lot about computing but little about their domain. But by learning a little about computing technology, domain specialists can now create tools that better embody their domain knowledge. Instead of compromising with a few centralized tools doing a mediocre job at many different tasks, domain experts can maintain a smaller collection of personal tools that do an excellent job of accomplishing their most important tasks. And when they have a new idea they’d like to try, they can do so immediately, rather than hoping that some faraway application developer will notice, agree with, and implement their suggestion. Good ideas can be shared easily among colleagues and migrated smoothly between different educational contexts.
One possible future is filled with useful and innovative technology built by users—the same people who have always been the experts at how the technology they use should work. That vision is made possible by coding agents, generative AI, and the technology companies developing these tools. But that same future is deeply threatening to the large part of the technology industry whose profits depend on users who can’t be builders.
Danny Hillis has described this possible future as “tech without the tech companies”, but it’s probably more accurate to say that it represents tech without most of the tech companies. No longer a monolith, the industry is splitting in two. Some tech companies are positioning themselves for this future—certainly those creating and deploying the models that support agentic development, but also new ones that will emerge to create new tools and environments for the new generation of programmers. But others have no choice but to try to preserve the status quo as long as possible. They’ve counted the chairs and decided that the music must continue.
Educators and other domain specialists can help bring about a future where technology is created and maintained by those who care about it. Educators are an intelligent and resourceful bunch. John, Ben, and many other teachers are already demonstrating what they can do with these new tools: motivated by their desire to support student learning, and guided by their domain expertise. There are certainly barriers ahead. Educational institutions are unprepared for educators who can build: lacking the required computation resources, budget lines, and institutional support. Educational technology companies will marshal ever more desperate arguments in support of their own survival, and may pursue regulations intended to make small educational tools less viable. Change isn’t inevitable—at least not on appropriate timelines. It’s going to require effort and advocacy.
University educators can help lead the way. Compared with most K–12 educators, we have a great deal of pedagogical freedom. Few university faculty are actually forced to use poor educational tools. Institutions may encourage, and faculty may go along, but outright coercion is rare. Imagine firing an instructor for creating superior educational technology! The biggest barrier here is institutional support for faculty-created educational tools—particularly funding for AI inference, which is urgently needed.
Computing educators may have the greatest opportunity of all. We can learn these tools ourselves and demonstrate to our colleagues how they are improving our courses. We can help colleagues from outside computing gain comfort and familiarity with the new computing—and then sit back and watch in awe as they transform education. And, most importantly, we can teach the new computing to a new generation of builders, abandoning the software priesthood in pursuit of tech without (most of) the tech companies.