Incoherent Thoughts from the Jagged Edge

[originally posted on my substack] 

I started this week sitting across from a long-time colleague, someone whose work I respect and whose judgment I trust. Let’s call him Eric. Eric is co-authoring something with a younger scholar, and he has been finding hallucinated references in the manuscript. He has, I think, tacitly accepted that AI was used in the writing. The references are the tell. He has not said so explicitly, but the evidence is sitting right there in the reference list.

What struck me about the conversation was not the hallucinations. Those are a real problem, and anyone using AI-assisted writing needs to know that. What struck me was that Eric is not using AI himself, which means his mental model of the technology is frozen at whatever he last heard about it. He does not know yet that frontier models (the current generation) are substantially less prone to fabricating citations than earlier versions. His experience of the technology is secondhand and dated, and that gap between perception and current reality is itself a kind of problem. The tool is a moving target, and the critique needs to move with it.

Later in the week, I had a completely different kind of conversation. Dave Fowler, a math education professor who retired a few years ago, shared with me the exchanges he has been having with ChatGPT about the nature of theories. Dave is curious, playful even about what this thing is and what it can do.

He shared some of the thinking he was doing with and about AI. We smirked together at the turns of phrase the model deployed in those conversations, including things like “You’re drawing a very precise and interesting analogy, Dave.” which my brain heard in the voice of HAL 9000. There is something both charming and slightly unnerving about that sentence. It flatters, it redirects, it sounds like a thoughtful interlocutor. Is it? Is it not? Dave was not sure, and neither was I, and that wondering felt like the right place to be.

I am still turning that conversation over in my head. What Dave is modeling is something I think we are not talking about enough in education: what it looks like to approach AI with curiosity rather than with a verdict already in hand. He retired, he has no institutional pressure either way, and for $20 a month he is just... exploring. There is something intellectually honest about that.

Uncertainty Saloon (Created with ChatGPT)

Meanwhile, the news cycle this week offered its own jagged edges.

A winner of the 2026 Commonwealth Short Story Prize has been accused of submitting an AI-generated story, and the situation quickly became a full-blown literary scandal. The accusations were amplified when Pangram Labs ran every Commonwealth Prize winner back to 2012 through their detection tool and found strong signals of AI-generated text in multiple entries, including the 2025 overall winner. The literary world, predictably, is having a moment.

And then Barnes & Noble CEO James Daunt appeared on the Today show and said he had no problem stocking AI-written books, as long as they were clearly labeled and not misrepresenting themselves. Cue the boycott calls. Cue the “all generative AI is ripping off someone else” counter-arguments on social media. Daunt has since clarified, repeatedly, that Barnes & Noble does not knowingly sell AI-generated books and takes active measures to exclude them. The clarification landed with about as much impact as you would expect, which is to say, not much. The outrage had already found its shape.

These two stories (the prize and the bookstore) are related. Both are really about thresholds. At what point does AI-assisted become AI-generated? Who gets to decide? What is the meaningful distinction between a human writer who uses AI as a tool and one who uses it as a ghostwriter? I do not have clean answers. What I am sure of is that any calls for no AI will just feed into shadow AI use.

On a more personal note: I have been noticing that when I write text on my own, without AI assistance, Grammarly flags it as AI-generated. (I still have a Grammarly subscription, though the controversy around the company has me reconsidering that). The flag itself made me stop and think. Am I absorbing AI patterns from all this use? Or is this a simpler and more uncomfortable explanation: that AI was trained on enormous quantities of mediocre writing (including mine), and is now reflecting the patterns of mediocre writing back at us, and Grammarly is simply recognizing them? I am a mediocre writer, so am I no better than AI?

I find this genuinely interesting and only slightly humbling.

Social media, that well-known venue for nuanced and measured discourse, has been full this week of a particular kind of certainty. The message, in various forms, is this: In five years you will look back and realize that AI in schools was a terrible mistake.

I do not know how to evaluate that claim, because I do not think anyone can know it. We are, as a society on the verge of something, but the outcomes are unclear. The people making this prediction with the most confidence tend to be the ones least encumbered by curiosity or evidence. As Ted Lasso misquoted “Be curious, not judgmental.” I think of Dave Fowler, sitting with his ChatGPT transcripts, genuinely wondering, and then I think of the social media certainty crowd, the UnDaves. The UnDave has an opinion. The UnDave does not let curiosity or facts complicate the opinion. The UnDave is very confident about what five years from now will look like.

I am not an UnDave. I hope I never become one. Here is where I am, for this week at least.

I love the exploration. I love what AI does for the scope and scale of what I can get done. And more than either of those things, I love that GenAI keeps opening doors to new questions, questions I would not have thought to ask, problems I would not have noticed, conversations like the one with Dave that I will be thinking about for weeks.

On schools specifically: I have said this before, and I will say it again. Caution is warranted. We do not have enough evidence yet about long-term impacts on learning, on writing development, and on productive struggle that builds capacity. Those concerns are legitimate and deserve to be taken seriously.

But on the other side of that, we must teach students about AI, what it is, how it works, and how it is changing the world they are growing up in. Not doing so is irresponsible. These students are going to live in a world shaped by this technology, whether we prepare them for it or not. Much like sex ed, if we don’t teach about it in schools, they will learn it elsewhere, with potential negative consequences.

I understand the desire to put the djinni back in the bottle. But we learned from the nuclear age that you cannot undo technologies. We must work as a society to reckon with the age of AI, and part of that is education.

Eric’s colleague used AI and did not know how to use it well. Dave used AI and understood enough to find the conversation genuinely interesting. The difference is not access. The difference is curiosity and intellectual engagement.

That is what education is (or at least should be).

Welcome to Summer!

 Summer School (The Good Kind)

Every year around this time, teachers start doing that particular mental math that only educators understand. How many days left? How many grades to enter? How many “are we doing anything today?” questions can one human absorb before the final bell? Summer is close, and the collective exhale is almost audible from here.

Most teachers I know fill that summer with a genuinely good mix of things: a conference or two, some professional reading they actually want to do, travel, family, and the radical act of sleeping past 6 AM. All of that is necessary and right. But I want to make a small case for adding one more thing to the list, and I promise it does not involve a syllabus or a rubric.

Get curious about GenAI.

Image Created with ChatGPT

Not in a high-stakes, district-mandate, someone-is-watching kind of way. Just curious. Summer is exactly the right time for this because the stakes are low. Nothing you try will scramble your fall classes. No 7:45 class depends on the quality of the output. You can experiment, fail, laugh at the results, and try again. That is almost never the condition under which teachers get to learn anything, and it is honestly the best condition for learning there is. Actually, the work we have done on ArtTEAMS grant for four years did exactly that each summer- but that is a story for another time. Although if you are interested our first publication just came out here.

Here are four things worth trying. These are just examples, feel free to riff…

Punch up an assignment you never had time to fix. We all carry them. The assignment that is fine, technically, but has been quietly bothering you for years because you know it could be better. Give a GenAI model the context, the learning objectives, and what you actually want students to get out of it, then see what it generates. I did this recently with an assignment where I always wanted students to think about instructional grouping using both assessment data and social and behavioral dynamics. I never had time to write meaningful student descriptions that would give the grouping decision real texture. This year I used GenAI to create those descriptions, and the assignment became substantially richer for it. The model did not replace my thinking. It handled the production work that had always been the obstacle.

Write a subject-specific AI policy. Counterintuitively, GenAI is quite good at helping you write a policy about GenAI. Start by feeding the model your district or institution’s existing policy, then describe the specificity and intentionality you want. I took this a step further and created differentiated expectations by assignment type: some asked students not to use AI at all, some encouraged proactive use, and some allowed it with disclosure. The clarity that came out of that process saved real class time and preempted a lot of the ambiguity that tends to eat up energy in the first weeks of a semester.

Make a document or presentation. The productivity tools built into or adjacent to GenAI have gotten genuinely good. I recently used one to produce a one-page project explainer that came out polished enough that I only needed to read and edit rather than draft from scratch. You still have to read it carefully and make it yours. But production time drops significantly, and for teachers who spend hours on materials that students may glance at for forty-five seconds, that math is worth considering.

Build a piece of software. This one surprises people, but it should not. GenAI is remarkably capable of creating small, functional, classroom-specific applications. I built an HTML-based reading fluency tracker that listens to a student read and another that scaffolds phonics practice. Both came out well beyond what I could have produced working alone, and neither required me to know how to code. The tools you actually want for your specific students, in your specific context, are more within reach than most teachers realize.

None of this requires a workshop. None of it needs to be assessed or reported. The only goal is to get your hands on the tools in a context where playing around is the whole point. That is how you develop genuine intuition about what these tools can and cannot do, which turns out to be far more useful than any training session.

So yes, go to the conference. Read the book. Take the trip. Sleep in. Spend lots of time with family and friends. Grill often. Drink occasionally.

And somewhere in there, try a few things. No pressure, no deadlines, no one watching. Just you, a curious question, and a summer with enough room to find out what happens. Or not, after all, it is up to you.

AI, Privacy, and the Context Conundrum

Something interesting happened recently in a conversation with Claude. I had been using a series of prompts recommended by Daniel Pink to do a kind of personal audit, and based on those conversations, I made some genuine changes. But I also noticed something that gave me pause.

Claude concluded that I was spending way too much time on administrative tasks and not enough on creative and research work. And while there is probably a kernel of truth in that, it was not quite right. The reality is that I lean heavily on AI for administrative tasks, and far less so for research and creative work, where most of my thinking happens in conversation with colleagues, on walks, or just away from the screen. Claude cannot see that work. What it can see is how I use Claude.

In other words, Claude was making inferences about my whole professional life based on how I have been using Claude. It reminded me of something I tell my students: they assume that because I teach, most of my time must go to teaching. In reality, it is about 40%. The AI was making the same natural, but limited, assumption. It was seeing the visible part of an iceberg and mapping the whole thing.
That was a useful insight on its own. But it pointed somewhere more interesting.

I recently listened to a discussion on the AI in Education podcast about bias in AI grading systems. One recommendation was straightforward: reduce the contextual information you give the AI about students. Remove names, gender, ethnicity. Strip away the signals that could activate bias. The less context, the less opportunity for those patterns to distort the evaluation.

That logic applies to me, too. The less context Claude has about me, the less it can stereotype or misread my work patterns. But here is where the conundrum arrives.

Context is precisely what makes AI more helpful.

Take a concrete example. Let me say, hypothetically, that I have a medical condition that makes me significantly less effective between 3 and 5 PM. If I want AI to help me plan my work week strategically, knowing that fact would make a real difference. It could help me schedule demanding intellectual work for the morning and reserve lighter tasks for those two hours. Without that context, I am just getting generic planning advice.But the moment I share that, I have handed a piece of genuinely private health information to an AI system, and by extension, to the company behind it. I may have no idea how that data is used, stored, or surfaced in future interactions. I have optimized for utility at the cost of privacy.
This is the lesson we already learned the hard way with social media. Early location-sharing felt like a fun, low-stakes way to connect. Foursquare check-ins were charming until they weren’t. The lure of personalization is real. The cost is often invisible until it isn’t. We traded something for convenience, and many of us are still sorting out what exactly we gave away.

For our own data, adults get to make that call. It is a tradeoff, and reasonable people will land in different places depending on their values, their risk tolerance, and how much they trust the platforms they use.

But student data is not ours to trade.

This is where I want to be unequivocal. The legal frameworks around student data, FERPA in the United States among them, exist for good reasons. Student data belongs to students and their families. When we use AI tools in educational settings, we are not making personal decisions about our own information. We are making decisions about children and young people who have not consented, who may not fully understand the implications, and who deserve protection.

So the practical guidance here is not subtle. Use only systems that are legally and contractually committed to protecting student data. Minimize the information you expose, even when a tool feels helpful. Resist the temptation of a quick AI fix that requires feeding it student names, identifiers, or demographic information.

The conundrum for adults using AI tools is real and worth sitting with. The tradeoff between context and privacy is genuinely complex.
For students, it is not a conundrum at all. It’s a responsibility.