Is Vibe Coding Changing Schools?

A podcast for educators on why non-technical school leaders are building AI tools in 2026, how vibe coding changes experimentation, and what this means for leadership, assessment, and edtech.

Transcript

SPEAKER_00 0:00

If this episode makes you think, please let us know in the comments and support us by subscribing and leaving a review. Thank you. Today we are exploring a really interesting article about school leaders building their own AI tools drawn from a piece describing a 10-week leadership cohort run with Sabah Quidwai, where non-technical education leaders are using AI, design thinking and vibe coding to solve real school problems. The article centers on examples from school systems across the US, Europe and Southern Africa, and argues that in 2026 the technical floor has dropped so low that leaders who do not think of themselves as technical are now shipping working prototypes in weeks, not months. And I have to say, this one really grabbed me, because on the surface this is a story about building tools. But underneath it, I think it is actually a story about permission. Permission to make, permission to test, permission to stop waiting for somebody else to solve your problem. According to the piece, these leaders are not engineers, they are deputy principals, assistant superintendents, innovation leaders, teachers, consultants, people rooted in real educational contexts, and instead of taking another course about AI in the abstract, they are spending ten weeks designing and shipping a real solution to a real problem in their own school or district. Three weeks in, most already had a working prototype. Now just sit with that for a second. How many times in education have we been told innovation is important, but the actual route to innovation has been painfully slow? You identify a problem, you form a committee, you go to procurement, you sit through demos, you trial a tool built for a market, not for your context, you adapt your practice around the software, and by the time anything lands the original problem has changed shape. What this piece suggests is that something fundamental has shifted. Not because schools suddenly became technical, but because the act of building has become conversational. The article explains vibe coding as a workflow where a human describes what software should do in plain English, and an AI agent builds it. It points to tools like Replit, Google AI Studio, Lovable and Claude making that possible for people without coding backgrounds. One example in the piece is especially striking. A prompt to build an AI policy assistant for a school trained on a PDF, and within five minutes there was a working branded chatbot on screen, ready for testing. Now I know what some people will be thinking. That sounds exciting but also a bit dangerous, a bit messy, a bit too easy. And yes, that reaction makes sense, but I think the more important question is this. What happens when the people who understand the daily friction of school life can finally build around that friction directly? That is the real shift here. According to the article, one leader in the cohort had previously seen herself as the person who came back from conferences full of apps nobody used. Another worried she simply did not have the skills to build what she was imagining. Both now have functioning prototypes teachers could click on. That is not just a technical update, that is a change in identity, the move from consumer to creator, from adopter to designer. And for schools that matters enormously. Because for years a huge amount of ed tech has asked schools to fit themselves around generic categories assessment platform, dashboard, parent communication layer, personalized learning system, policy assistant. But schools are not generic places. Their cultures are different, their constraints are different, their policies are different, their politics are different, their people are different. What works beautifully in one trust district or independent school can feel completely dead on arrival somewhere else. The article quotes Karen Brennan from Harvard talking about the democratization of creation and the economics of experimentation. I think that phrase is brilliant, the economics of experimentation, because the issue in schools has never just been imagination. Schools are full of imaginative people. The issue has been the cost of turning imagination into something testable. Time cost, technical cost, procurement cost, political cost, confidence cost. When those costs fall, behavior changes. And the examples in the article really show that. A district AI governance hub that pulls together collective bargaining agreements, strategic plans, learner vision documents and accreditation reports into one living system. A confidential coaching partner for teachers, designed not because coaching is bad, but because in that district coaching had become politically loaded, a governance assistant for small school boards with rotating parent volunteers, a student feedback tool that turns survey voice into actionable next steps for teachers within days instead of terms. And one that really stood out to me, a proof of thought protocol built to credit cognitive process over final product in assessment. That is where this gets especially interesting for educators. Notice what these leaders are not doing. They are not trying to build the next giant platform. They are not trying to recreate a student information system. They are not trying to disrupt education in the tired Silicon Valley sense. According to the article, they are solving sharply defined internal problems, and those problems are so context specific that off-the-shelf vendors often cannot really solve them well. That to me is a very important lesson. Start with why, not how, or to put it another way, purpose over technology. This is one of the big traps in AI conversations in education. We start with capability. Look what the model can do. Look what the tool can generate. Look at the slick demo. But the better question is always, what actual friction point are we addressing? What is persistently clunky, overloaded, delayed, fragmented, or too politically difficult in our current system? Because if you start there, the tool becomes a response to a real need, not another novelty, and that is exactly what the strongest examples in the piece seem to do. Take that coaching tool. I found that example especially revealing. According to the article, the assistant superintendent is not building it to replace human coaches, he is building it because human coaching in that context carries stigma from years of being associated with underperformance. So the AI tool becomes a low stakes re-entry point into reflection and may eventually lead more teachers back toward working with the human coaches already in place. That is such a small educational design move. It is not replacement, it is relational strategy. And that is where I think a lot of shallow AI commentary misses the point. The best users are often not about automation alone. They are about reducing friction around the human work we actually value. Outsource the doing, not the thinking. Outsource the awkward first step maybe. Outsource the admin drag. Outsource the sorting and synthesis. But keep the judgment, the care, the interpretation, the relationship. The article says much the same later on. The leadership and most consistently are not the ones most excited about AI. They are the clearest about the human work they are trying to protect. That line stayed with me, because in schools the real danger is not that AI exists, the real danger is that we use it lazily, that we allow convenience to flatten thought, that we chase polished outputs instead of stronger learning, that we reduce teacher professionalism to prompt copying. But when school leaders are clear that the purpose is to protect human judgment, teacher confidence, student voice, board quality or more meaningful assessment, then AI becomes something very different. Not a replacement machine, a design material. And the assessment example in the article opens up a huge question for schools, a proof of thought protocol that credits the cognitive process over the final product. Partly in response to cheating panic, but also because, according to the piece, the panic was always pointing to a deeper problem about what schools actually assess. Exactly. This is the conversation we need to be having. Not just how do we stop students using AI, but what kinds of evidence of learning still matter in an age when polished product is easier to generate? This is where assessment needs redesign. Product still matters, yes, but so do process and performance. How did the student get there? What decisions did they make? What did they discard? What can they explain live? What can they defend? What can they adapt? Design learning that cannot be faked because it demands depth, care and imagination. And that applies to leaders too, by the way. Because another strong theme in the piece is narrowness. According to the article, almost every leader began with an idea two or three times bigger than what they eventually shipped, and the breakthroughs came when they cut the scope down. Small experiments, math only, not whole school. One board first, not dozens of schools, one pilot before scale. There is so much wisdom in that. Education often swings between two bad habits, either tiny isolated pilots that never go anywhere, or giant strategic ambitions that collapse under their own weight. What this article points toward is something more useful. Small enough to build, real enough to matter, specific enough to test. That is the economics of experimentation in action. The second pattern the article highlights is feedback loops over launch events. Again, brilliant. The most effective leaders, it says, are not unveiling polished tools at staff meetings. They are getting rough prototypes in front of real users quickly, including skeptical users, and letting those people shape what comes next. Honestly, schools need much more of that mindset. Not because everything should be beta forever, but because so much school change is overproduced and undertested. We spend months writing documents no one reads, then wonder why implementation fails. But a rough prototype in front of 150 students, a live tool tested by the very teachers it is meant to support, a governance assistant piloted by the actual board that needs it. That is much healthier, more honest, more human, more adaptive. And then there is the wider signal in the article. It points to Peninsula School District in Washington expecting to drop three to four software subscriptions because it has vibe-coded its own replacements. It also notes a US Department of Education rule giving priority to AI implementation efforts in discretionary federal grants, suggesting districts that can demonstrate built capability may move ahead of those that cannot. Now I want to be careful with this part because one article is not a full market analysis, but the strategic implication is obvious enough. If districts begin cancelling several generic tools because they can build tailored alternatives themselves, then parts of the ed tech market become much shakier. The article makes exactly that point. Systems with deep compliance or regulated functions may be fine. Generic productivity layers and thin AI wrappers may be much more vulnerable, and for school leaders that should prompt a really healthy question. What do we truly need to buy? And what might we increasingly be able to build? Not everything, obviously. Schools should not become accidental software companies. That would be absurd, but neither should they assume that every persistent problem requires a vendor. Some problems need a platform, some need a policy, some need staff development, some need a workflow redesign, and some, increasingly, may need a small, context-rich tool built by the people who understand the problem best. That is a very different leadership model. It means the future facing school leader may need less confidence in technology features and more confidence in problem framing, less obsession with tools, more clarity about workflow, less what app should we buy, and more what friction are we trying to remove? Less top-down digital strategy as document, more strategy as making and testing. And there is something quite hopeful in that. Because one of the quietest but most powerful claims in the article comes right at the end. For the first time, the people who know education best can build the tools they need without permission, procurement, or a six-month implementation cycle. That does not mean all tools built this way will be good, of course not. Some will be clumsy, some will fail, some will solve the wrong problem, but that is not the point. The point is that capability has moved closer to context, and when capability moves closer to context, innovation gets more honest. So for me, the educational takeaway here is not everyone should start vibe coding tomorrow. That would miss the point entirely. The deeper lesson is that school leadership in the AI era may be becoming more design oriented, more experimental, more local, more iterative, less reliant on waiting for perfect systems to arrive from elsewhere. And perhaps most importantly, the leaders who will do this best are not the most technical. According to the article, they are the ones who are clearest about the human work they are trying to protect and the narrow problem they are trying to solve. That feels exactly right to me because machines can compute. They cannot wonder, they cannot care. The role of the school leader is not to become the coder in chief. It is to notice where the system is wasting human energy, where professional judgment is being blocked by bureaucracy, where student voice is trapped in dead documents, where assessment no longer matches learning, and where a small, well built tool could create new room for human agency. That is a very exciting prospect. Not because it makes schools more technical, but because it may finally allow them to be more themselves. That's all for today. Thanks for listening.