In June 2025 I published Fear Is Not a Framework, a piece that argued the panic around AI in education was missing the point. I framed cognitive offloading as a feature of human intelligence, not a flaw. I argued that the right response to ChatGPT in schools was design and pedagogy, not refusal. I offered a five-principle pedagogical framework and a specific assignment model: draft with AI, then revise, then reflect. I made the case in good faith. I believed it.
In April 2026 I published Show Your Work, a three-part essay that argues almost the opposite. The evidence for AI in U.S. K-12 classrooms does not exist where adoption is happening. The historical analogy to calculators and word processors does not hold for tools that perform the cognitive task itself. The right of educators and students to refuse these tools is something I now think we have to defend.
Someone reading both will notice the contradiction. They should. I owe an account of it.
Here is what I want to say about June 2025. It was true in its moment. The AI conversation a year ago was a real conversation. People were asking real questions. The studies that anchor Show Your Work did not exist yet. SCALE was eight months away. The Wharton/PNAS RCT had not been published. Gartner's forecast on AI-free skills assessments had not been issued. The Common Sense Media data on teen chatbot use had not landed. I was writing in a moment when "build thoughtful pedagogy around this tool" looked like the responsible centrist position. I held it because I thought it was right.
Two things changed.
The first is that the evidence accumulated. I won't repeat what I laid out in Show Your Work. It is there for anyone who wants to read it. What matters is that the empirical picture in early 2026 is not the picture from mid-2025, and I am not interested in defending a 2025 argument with 2026 data. Sweden looked at its own PISA scores and reversed a digital-first policy that had once been a national point of pride. That is the move available to anyone honest enough to make it. I am trying to make it here.
The second is harder to write. The educational conversation about AI did not develop into what Fear Is Not a Framework was arguing for. It calcified into something else. The thoughtful integration model I described, with iteration and reflection and explicit scaffolding, is not what is happening in most schools. What is happening is adoption, full stop. Conference programs that present AI as the answer to every question already on the agenda. Vendors offering free products with year-four catches nobody is asking about. Teacher burnout addressed not by lowering expectations but by letting a chatbot generate "real-sounding" comments on student work. Research instruction reframed as prompt engineering. The conversation about AI in education stopped being a conversation. It became, in tone and structure, something closer to a faith. The people inside it speak in the language of conversion. The people outside it are treated as having not yet seen the light.
I am not religious. I notice when something is being sold to me with the rhythm of religion.
The 2025 prescription assumed conditions that mostly do not exist. It assumed teachers had the time to design iterative assignments with reflection components. It assumed administrators would let "the right to refuse" stand as a professional position rather than a problem to be managed. It assumed students approached AI from a stance of curiosity rather than from the stance any kid would reasonably take, which is: this tool can make school easier, why wouldn't I use it. That last point is the one I had not properly weighed. I was writing pedagogy for an imagined classroom. I should have been writing for the actual one.
This year I have started seeing something in our library I did not see before. Kids using AI with a posture I can only describe as furtive. The shoulder-check. The "who is watching" glance. And the faculty around them being told, often by the same administration, that they should be promoting and modeling AI use. The kids are hiding. The adults are evangelizing. Nobody seems to find this strange. I do.
I have watched freshmen and sophomores complete long-form research projects and then fail to explain the gaps in their own research. You do not need a confession to know what happened. You only need to have taught research long enough to recognize the shape of a project no one actually built.
So here is what I now think I got wrong in June 2025. I underestimated how completely "use the tool well" requires conditions of time, scaffolding, professional judgment, and student development that are not present in most schools and are not coming. I overestimated how much the educational AI conversation would resist the gravitational pull of the marketing around it. I framed the refusal of these tools as a less serious position than the thoughtful-integration position. It is not. In the conditions we actually have, refusal is one of the most serious responses available to a professional educator, and I want to defend the people taking it rather than write essays that quietly make their position harder to hold.
One thing I will not give back. I still use AI a lot. I build apps with it. I draft documents with it. I think about problems with it. That is not a contradiction. I am a forty-something professional with developed skills, working judgment, and the metacognitive equipment to know when a tool is doing the thinking for me. The kids I work with are building that equipment right now. The whole point of school is that they get to build it. A tool that performs the task for them while they are supposed to be learning the task is not the same tool I use to draft a curriculum doc, even if the interface is identical.
Fear Is Not a Framework said that friction is sometimes just friction. I want to revise that. For an expert, sometimes. For a fourteen-year-old learning to think, friction is where the thinking happens. Without it, there is no learning. There is only output.
I left Fear Is Not a Framework up. I am not interested in pretending I did not write it. It is the record of where I was a year ago. Show Your Work is the record of where I am now. This piece is the bridge, and the bridge is the honest thing.