Hi — I taught math in NYC schools and now private tutor while building an AI math tutor for my students. Your post touches on real struggles teachers are about to face, and I agree with most of it. I'd push back on one piece though.
For me this hasn't really been about learning to "think like a coder." It's been about tolerating being wrong over and over while you figure out what actually works with AI. You build, you test, you realize you were thinking about it wrong, you start over. That's a teacher skill — we do it with lessons all the time.
What worries me more than the coder gap is the time gap. Teachers don't have evenings and weekends to spend on this the way solo builders do. That's the real wall, and I don't think anyone in the industry has answered it yet.
On the coding, a lot of it is the jargon I'm starting to encounter - but the kind of thinking your describing is "thinking like a coder" and is quite useful - it's just more technical than I'm used to. In any case, you've identified probably the most important piece - a lot of this experimentation is time-intensive and that's one thing many teachers don't have.
Exactly — and I think this is where teachers helping each other matters more than usual. Most of what I’ve learned about working with AI didn’t come from documentation. It came from talking to other educators, figuring out what they wanted the AI to do, and working through it together until it actually did the thing. That’s the community I’m trying to build, slowly — educators learning and helping each other. If we don’t make the time, no one’s going to make it for us.
Yet I'm curious about the lens of "humanities teachers needing to get on board," as it feels like a much broader expertise gap—one that, at least for me, feels more and more divorced from the parameters of a typical high school English classroom.
Yes: I think we should continue to be offering opportunities for digital literacy, introductions into coding, etc. If I had my way, this would be a core sequence through high school alongside Math, Science, ELA, and Social Studies.
But: I guess I don't feel obligated at all to set aside time as an ELA teacher to prioritize this right now, particularly when it comes to the coding language/requirements for these new tiers of use.
I don't think I say they need to "get on board" - my point is that for those who do (or want to pursue more AI literacy), it's going to continue to be a steeper learning curve. On the plus side, I don't thing the vast majority of students are close to this - but if teachers ignore it, it will further widen the gap. But I tend to agree on the way to get at it - I do think it's going to require a stand alone course that seems like it should reside in computer science. Awareness is key, though, especially as there will always be some kids doing interesting things with AI - should they be ignored, chastised, or supported? And to be clear, the whole point is you don't actually know how to code, but the coding mindset is taking hold which has all sorts of pluses and minuses. Anyway, that's what I'm noticing!
Oh, agree with all of this! And part of this could have been me leaning too much on the sub-header and reading all of this before my coffee fully kicked in 😂
For me, in some ways it feels like going full-circle to where we were when "coding was the cool thing!"—two things were true then and I think are true now: (1) there's a ton of value for those who can master that skillset and we should be providing that value as much as possible in schools and (2) there's not an implicit pressure on all teachers to assume responsibility into their content areas.
The conceptual complexity you're experiencing has a structural explanation: putting an LLM in a harness doesn't create an agent, it creates a workflow with an unreliable reasoner. The complexity is what you accumulate engineering around that gap—each layer of orchestration is another attempt to compensate for a component that can't reliably do what the marketing implies.
Excellent essay Stephen that hints at a major shift happening in “AI world” as the major companies look to monetize their products. Nilay Patel recently published an insightful essay on “software brain” that dovetails nicely with yours. In his words:
“So what is software brain? The simplest definition I’ve come up with is that it’s when you see the whole world as a series of databases that can be controlled with the structured language of software code….But: not everything is a business. Not everything is a loop! The entire human experience cannot be captured in a database. That’s the limit of software brain. That’s why people hate AI. It flattens them.”
“The chatbot is now the floor” honestly feels like the line a lot of educators are going to keep coming back to. I think many people still picture AI as asking questions in a text box, while the expectations around actual use are already shifting underneath them.
Your ending ("When I spent a summer as a paralegal, one of the partners explained to me that every profession has its own secret vocabulary, its own language that sets it apart.") made me wonder about the secret language of generations apart from professions.
Perhaps students will learn to use 2032 AI tools without the professional language beyond "vibe coding" and they attach to those technical terms the 'skibidy' and 'riz' of new generations. It will be interesting to see language morph with tech.
"Withholding AI from students until they’re 'ready' assumes a static technology which currently doesn’t exist."
Nor will it ever.
Students' understanding of motivations, patterns, and structural "thinking" in machines is not just a sensible complement to their invaluable humanities education. It's quite necessary.
You’re rightStephen — I’m a high school chemistry and physics teacher. Not a coder. And I use AI every single day. Writing books, building articles, creating simulations, automating workflows, making videos. The tools you’re describing aren’t theoretical for me. I’ve crossed that threshold you’re talking about, and I think I can name what got me there.
It wasn’t learning to think like a coder. It was having the internal skills already in place — the ability to hold a problem in my head, break it into parts, test my own thinking, and know when I’m being handed a design choice versus a finished product. Those are the cognitive skills that make AI usable at the level you’re describing. And they’re the same skills that existed long before any of this technology did.
Here’s what I see in my classroom that connects to your argument: my students can’t do this stuff. Not because they won’t try — they will. It’s because they came up through COVID, and the foundational cognitive habits that would let them engage with these tools never fully developed. The internal operating system got disrupted. They struggle to hold complexity, tolerate ambiguity, and sequence their own thinking without external scaffolding. That’s not a technology problem. It’s a cognitive one. I wrote about this extensively in my book Cognitive Sovereignty Under Compression — the idea that learning to think is a non-compressible process, and that no tool, no matter how powerful, can substitute for it.
Your “eternal skills” section gets closest to what I think the real answer is, but I’d push it further. It’s not that eternal skills will eventually produce AI fluency as a byproduct. It’s that without those internal capacities — what I call cognitive sovereignty — the tools are unreachable no matter how good the interface gets. The gap you’re identifying isn’t really between coders and non-coders. It’s between people who own their thinking and people who don’t. The interface will keep improving. The internal skills won’t build themselves.
That conviction is why I built Thinking Labs — a program that teaches students and adults how to develop those internal capacities before they ever touch a tool. Because everything you’re describing about the future of AI fluency depends on a foundation most of our students never got the chance to build — and that a lot of us as educators are still building in ourselves. Teachers by nature are lifelong learners. If we can teach kids to become lifelong learners too, AI fluency comes naturally.
Wonderful interpretation and insightful analysis. I am an independent career counselor guiding students from 9th grade onwards to help them choose the right career domain. I also do web development as part of my passion... Sometimes I find myself contradicting as I use AI for coding things out and then also talk about how AI is reducing human ability... :)
Withholding AI from students until they’re “ready” assumes a static technology which currently doesn’t exist.
Not necesssarily. It gives them time to learn, ie absorb content (yes, that still exists and is fundamental) which becomes internalised knowledge. Not to mention, if the AI is galloping at the pace you describe, by the time they leave school it will have reached a different dimension anyway.
Thanks, Susan. I might agree if it weren't for the fact that students are using AI constantly to help with and complete the very work we assign. There is no world I see in which they will not be using it themselves - the question to me is whether we let them discover it on their own or under our guidance. I come out on the latter.
I think this is completely right and I got a glimpse of this near future last year, when I attempted to self-learn how to use AI agents. I started watching YouTube videos that were an hour long(!) and felt my eyes glaze over. If it takes an hour (of sped up video) to explain how to use n8n, maybe it’s above my pay grade. I was hoping we’d see a platform for agents that felt as frictionless as a context window, but that moment hasn’t arrived yet. Instead we got an arcane tool like OpenClaw, definitely out of reach for the average vibe coder. Thanks for saying out loud what some of us have been witnessing for months!
good piece, thank you. It is crucial that schools support teachers in every way to follow this fast evolution, and what it means for students/teachers. This piece would also suggest that most tools by the Edtech industry will become obsolete ever faster, and schools need to take ownership of what they want to do with AI and how to do it, rather than paying for (questionably relevant) systems which will disappear as fast as they appeared. I don't think most of the schools are ready for that.
Hi — I taught math in NYC schools and now private tutor while building an AI math tutor for my students. Your post touches on real struggles teachers are about to face, and I agree with most of it. I'd push back on one piece though.
For me this hasn't really been about learning to "think like a coder." It's been about tolerating being wrong over and over while you figure out what actually works with AI. You build, you test, you realize you were thinking about it wrong, you start over. That's a teacher skill — we do it with lessons all the time.
What worries me more than the coder gap is the time gap. Teachers don't have evenings and weekends to spend on this the way solo builders do. That's the real wall, and I don't think anyone in the industry has answered it yet.
On the coding, a lot of it is the jargon I'm starting to encounter - but the kind of thinking your describing is "thinking like a coder" and is quite useful - it's just more technical than I'm used to. In any case, you've identified probably the most important piece - a lot of this experimentation is time-intensive and that's one thing many teachers don't have.
Exactly — and I think this is where teachers helping each other matters more than usual. Most of what I’ve learned about working with AI didn’t come from documentation. It came from talking to other educators, figuring out what they wanted the AI to do, and working through it together until it actually did the thing. That’s the community I’m trying to build, slowly — educators learning and helping each other. If we don’t make the time, no one’s going to make it for us.
All well-documented and, as usual, insightful!
Yet I'm curious about the lens of "humanities teachers needing to get on board," as it feels like a much broader expertise gap—one that, at least for me, feels more and more divorced from the parameters of a typical high school English classroom.
Yes: I think we should continue to be offering opportunities for digital literacy, introductions into coding, etc. If I had my way, this would be a core sequence through high school alongside Math, Science, ELA, and Social Studies.
But: I guess I don't feel obligated at all to set aside time as an ELA teacher to prioritize this right now, particularly when it comes to the coding language/requirements for these new tiers of use.
I don't think I say they need to "get on board" - my point is that for those who do (or want to pursue more AI literacy), it's going to continue to be a steeper learning curve. On the plus side, I don't thing the vast majority of students are close to this - but if teachers ignore it, it will further widen the gap. But I tend to agree on the way to get at it - I do think it's going to require a stand alone course that seems like it should reside in computer science. Awareness is key, though, especially as there will always be some kids doing interesting things with AI - should they be ignored, chastised, or supported? And to be clear, the whole point is you don't actually know how to code, but the coding mindset is taking hold which has all sorts of pluses and minuses. Anyway, that's what I'm noticing!
Oh, agree with all of this! And part of this could have been me leaning too much on the sub-header and reading all of this before my coffee fully kicked in 😂
For me, in some ways it feels like going full-circle to where we were when "coding was the cool thing!"—two things were true then and I think are true now: (1) there's a ton of value for those who can master that skillset and we should be providing that value as much as possible in schools and (2) there's not an implicit pressure on all teachers to assume responsibility into their content areas.
(And yes to a computer science core path!)
The conceptual complexity you're experiencing has a structural explanation: putting an LLM in a harness doesn't create an agent, it creates a workflow with an unreliable reasoner. The complexity is what you accumulate engineering around that gap—each layer of orchestration is another attempt to compensate for a component that can't reliably do what the marketing implies.
I explored the security implications of this in December—why agentic AI can't be secured because the "agent" is a category error, not a design choice. Seems relevant to where you've landed: https://thepuzzleanditspieces.substack.com/p/the-agent-that-wasnt-there
Excellent essay Stephen that hints at a major shift happening in “AI world” as the major companies look to monetize their products. Nilay Patel recently published an insightful essay on “software brain” that dovetails nicely with yours. In his words:
“So what is software brain? The simplest definition I’ve come up with is that it’s when you see the whole world as a series of databases that can be controlled with the structured language of software code….But: not everything is a business. Not everything is a loop! The entire human experience cannot be captured in a database. That’s the limit of software brain. That’s why people hate AI. It flattens them.”
https://www.theverge.com/podcast/917029/software-brain-ai-backlash-databases-automation
“The chatbot is now the floor” honestly feels like the line a lot of educators are going to keep coming back to. I think many people still picture AI as asking questions in a text box, while the expectations around actual use are already shifting underneath them.
I have actually just created curriculum for middle schoolers and each lesson (of 12) has a section for the parent about any it matters: https://shapingmindsco.etsy.com/listing/4496329602
Your ending ("When I spent a summer as a paralegal, one of the partners explained to me that every profession has its own secret vocabulary, its own language that sets it apart.") made me wonder about the secret language of generations apart from professions.
Perhaps students will learn to use 2032 AI tools without the professional language beyond "vibe coding" and they attach to those technical terms the 'skibidy' and 'riz' of new generations. It will be interesting to see language morph with tech.
They may not even need to write it and just beam it from their brains!
The EEG machines are becoming less obtrusive. EEG through headphones...I can see it.
Your essay left me wondering: will students have to change to meet the needs of AI? Or will AI have to change to meet the needs of students?
"Withholding AI from students until they’re 'ready' assumes a static technology which currently doesn’t exist."
Nor will it ever.
Students' understanding of motivations, patterns, and structural "thinking" in machines is not just a sensible complement to their invaluable humanities education. It's quite necessary.
You’re rightStephen — I’m a high school chemistry and physics teacher. Not a coder. And I use AI every single day. Writing books, building articles, creating simulations, automating workflows, making videos. The tools you’re describing aren’t theoretical for me. I’ve crossed that threshold you’re talking about, and I think I can name what got me there.
It wasn’t learning to think like a coder. It was having the internal skills already in place — the ability to hold a problem in my head, break it into parts, test my own thinking, and know when I’m being handed a design choice versus a finished product. Those are the cognitive skills that make AI usable at the level you’re describing. And they’re the same skills that existed long before any of this technology did.
Here’s what I see in my classroom that connects to your argument: my students can’t do this stuff. Not because they won’t try — they will. It’s because they came up through COVID, and the foundational cognitive habits that would let them engage with these tools never fully developed. The internal operating system got disrupted. They struggle to hold complexity, tolerate ambiguity, and sequence their own thinking without external scaffolding. That’s not a technology problem. It’s a cognitive one. I wrote about this extensively in my book Cognitive Sovereignty Under Compression — the idea that learning to think is a non-compressible process, and that no tool, no matter how powerful, can substitute for it.
Your “eternal skills” section gets closest to what I think the real answer is, but I’d push it further. It’s not that eternal skills will eventually produce AI fluency as a byproduct. It’s that without those internal capacities — what I call cognitive sovereignty — the tools are unreachable no matter how good the interface gets. The gap you’re identifying isn’t really between coders and non-coders. It’s between people who own their thinking and people who don’t. The interface will keep improving. The internal skills won’t build themselves.
That conviction is why I built Thinking Labs — a program that teaches students and adults how to develop those internal capacities before they ever touch a tool. Because everything you’re describing about the future of AI fluency depends on a foundation most of our students never got the chance to build — and that a lot of us as educators are still building in ourselves. Teachers by nature are lifelong learners. If we can teach kids to become lifelong learners too, AI fluency comes naturally.
Wonderful interpretation and insightful analysis. I am an independent career counselor guiding students from 9th grade onwards to help them choose the right career domain. I also do web development as part of my passion... Sometimes I find myself contradicting as I use AI for coding things out and then also talk about how AI is reducing human ability... :)
Withholding AI from students until they’re “ready” assumes a static technology which currently doesn’t exist.
Not necesssarily. It gives them time to learn, ie absorb content (yes, that still exists and is fundamental) which becomes internalised knowledge. Not to mention, if the AI is galloping at the pace you describe, by the time they leave school it will have reached a different dimension anyway.
Thanks, Susan. I might agree if it weren't for the fact that students are using AI constantly to help with and complete the very work we assign. There is no world I see in which they will not be using it themselves - the question to me is whether we let them discover it on their own or under our guidance. I come out on the latter.
On these grounds I completely agree.
I think this is completely right and I got a glimpse of this near future last year, when I attempted to self-learn how to use AI agents. I started watching YouTube videos that were an hour long(!) and felt my eyes glaze over. If it takes an hour (of sped up video) to explain how to use n8n, maybe it’s above my pay grade. I was hoping we’d see a platform for agents that felt as frictionless as a context window, but that moment hasn’t arrived yet. Instead we got an arcane tool like OpenClaw, definitely out of reach for the average vibe coder. Thanks for saying out loud what some of us have been witnessing for months!
good piece, thank you. It is crucial that schools support teachers in every way to follow this fast evolution, and what it means for students/teachers. This piece would also suggest that most tools by the Edtech industry will become obsolete ever faster, and schools need to take ownership of what they want to do with AI and how to do it, rather than paying for (questionably relevant) systems which will disappear as fast as they appeared. I don't think most of the schools are ready for that.
Remember the “Smart Boards” …. shiver…..(: