One of my most vivid memories from my first years of teaching is a faculty meeting where we were discussing the introduction of email (yes, it was a long time ago) and what our responsibilities would be when it came to communicating with parents and administrators. I had just come from a career in law where our firm had underdone a similar transition. One of the longest-tenured teachers, a gentleman from the pre-digital age, simply sighed and said, “Can we just opt out?” He was hoping he could be exempt from having an email address at all. After close to forty years in a world where he was insulated from a daily barrage of interruptions, he was asking if he could continue without these intrusions in the twilight of his teaching life. Nearly thirty years later, present-day me sympathizes with those very prescient concerns. Of course, he did not get to opt out. None of us did.

I thought about him last week when I came across a short piece in the Chronicle of Higher Education about David Durman, a research administrator at the University of Kentucky. Durman, who identifies as a humanist, applied for and received a religious exemption from using AI at work. Though never required to use it, he wanted to pre-empt the possibility because his university had begun “suggesting employees incorporate AI into their workflows.”

“A component (of humanism) is the aspect of dignity to the human self, and with this technology, it’s different than perhaps a simple autocorrect in a document or some sort of assistive device like that, in that it’s the offloading of cognitive ability, the offloading of decision making, and so on,” he said. “You have to have ownership of your cognitive processes and your ability to determine your path to the world and abdicating that, or even parts of that, lessens the person.”

The University granted the accommodation.

Jay Blanton, a university spokesperson, said in a statement: “In this case, the university determined that an accommodation could be provided while allowing the employee to continue performing the essential functions of their position. As technology and job responsibilities evolve, accommodations may be reviewed to ensure they remain reasonable and workable.” (emphasis added)

Will a new hire in 2030 be offered one as well? They will have the same right to ask, but what is the likelihood that the job description for “research administrator” designates AI use as an essential function of their work, if that job even exists at all? Many companies, including my wife’s, are strongly encouraging employees to use AI. I also know people in industries where each employee receives a monthly allowance of AI tokens, usage is checked at the end of the month, and those metrics are, presumably, tied to performance reviews and bonuses. I don’t see how that kind of corporate structure will be compatible with anti-AI religious exemptions, and I can’t imagine such a stance is going to enhance career opportunities in many fields of knowledge work.

The administrative side of a school or university operates more like a business. Admissions, advancement, scheduling, and HR are all being offered AI tools in the name of efficiency, and institutions trying to manage their bottom lines more aggressively have every incentive to use them. While schools are typically slower to change than the private sector, as the financial squeeze across higher ed persists, if anything those incentives will increase. Despite the success of Durman’s request, I suspect it will be much, much tougher to opt out of using AI on the administrative side in a few years’ time, and the disconnect is difficult to square while the broader conversation about AI on campus remains so divisive.

Professional autonomy and the very real pedagogical concerns around learning make it likely that classroom AI practice will remain the most protected part of the job for much longer. Email didn’t change how we taught, but it did change the administrative infrastructure of how the job is done and all the ancillary requirements surrounding it. There was no “opting out.” If AI becomes “inevitable” for faculty, I would expect it to come through communication, reporting, learning management systems, and grade books, as those digital systems undergo the same overhaul we are seeing on the student side of edtech.

Schools are, if anything, more practiced at accommodation than most workplaces. As a teacher currently, and selectively, experimenting with AI in the classroom, I will always provide an opt-out for students who are uncomfortable with the technology. I have not had a student raise a religious objection, but I have no desire to force a student to use an AI tool if they do not wish to. Teachers are also routinely asked to make accommodations around technology, some of them contradictory, but each of which makes sense on their own terms. A student with a graphomotor diagnosis may use a laptop in class regardless of the teacher’s individual policy, while another student may be exempt from screens entirely while recovering from a concussion. I do wonder whether that same generosity will extend to the adults at school if and when AI becomes seamlessly integrated into the administrative layer. Right now, many faculty at schools rapidly adopting AI are undergoing the same experience as my older colleague did about email in the mid-90s. How much longer will they be able to opt out?

I find myself wondering what my current seniors will face when they enter a job market in less than five years. We have gone from clunky chatbots in early 2023 to AI agents that can take on real chunks of white-collar work, solve advanced math and physics problems, and write competent software from plain-English instructions as we near the end of 2026. What awaits us over the next 36 - 60 months?

One of my students told me recently that he thought people already in the workforce were in worse shape than his generation, because they built their careers before AI arrived, while he and his classmates will have grown up with it. I’m not sure he’s right and I don’t think that’s where many of the predictions are right now, but I empathize with how students are currently analyzing their career options.

We seem to be in this interregnum period where many AI issues currently being litigated - whether it’s the rate of improvement of the models, the role of AI in the economy, or the impact of AI on the workforce - will, if not be “resolved,” at least come into clearer focus as we get closer to 2030. That’s not to say we won’t still be arguing about AI, but either scaling will have slowed down, the AI bubble will have burst, and/or the labor market will show greater signs of disruption.

Despite the significant penetration of AI into society overall, a large number of people have never used one. Many teachers still want nothing to do with AI. They don’t want it near kids, and they aren’t playing or experimenting with it at all. I am genuinely curious whether that will still be possible four years from now. If you had asked me in early 2023 whether teachers would still be able to avoid AI entirely by the fall of 2026, I would have said no. I was clearly wrong.

But I do remember telling my older colleague that, when it came to email, it wasn’t a matter of “if” but “when.”

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions on pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

Though I’ve been a secondary school history teacher for over 30 years, my interests range across many subjects. I knew early on that I leaned more toward the humanities than STEM - my only foray into the sciences in college was Physics for Poets - but I always envied my scientifically inclined peers and instinctively understood how important those subjects were.

So when I read the coverage of OpenAI’s October 6 announcement that its newest model, one not yet released to the public, had produced hundreds of new mathematical results, I was intrigued. Back in May, the company said that one of its models had disproved a conjecture Paul Erdős posed in 1946, a result many top mathematicians accepted as significant. In early September, it claimed its models had solved Navier-Stokes, one of the seven official Millennium Prize Problems in mathematics.

The last two caused an uproar: Navier-Stokes over its timing and the lack of credit given to other mathematical researchers, and the October release over its sheer volume. That OpenAI courts controversy should surprise no one familiar with its track record. The Navier-Stokes claim is still being checked, and some mathematicians argue OpenAI solved a narrower version of the problem than the headlines suggested. Nevertheless, there seems to be broad acknowledgment that AI is producing genuinely new mathematics, and that recognition has engendered something like a collective mourning in the field.

The New York Times’ Daily podcast ran an illuminating interview with one such mathematician, Steven Strogatz of Cornell, on October 13th. What both impressed and depressed Strogatz was that the Erdős disproof was elegant, connecting two areas of math, discrete geometry and number theory. He called the connection “the magic key.” As I understood him, it impressed him because until then, most of the math produced by AI had seemed mechanical. It depressed him because … well, if AI proofs could be elegant, what did that mean for human mathematicians?

At one point, Strogatz gets choked up describing the famous 1985 photo of Erdős working with a pre-teen Terence Tao and his sense that this kind of work may disappear. On Friday, Tao told the Times he’s “on a war footing,” and said senior academics like him aren’t as threatened as those just starting out. One of his PhD students, Ben Goldman, said he was upset to see the field he loves being “strip mined.”1

Cheating Is Still the Lens Most Schools Are Fixated On

While most schools are still struggling with cheating policies, AI literacy, and other basic issues around unauthorized student use, the major companies seem to have bigger fish to fry. Few teachers likely understand, use, or appreciate the agentic features offered in some of the free tiers, and the paid tiers continue to pull away from the free models. The way OpenAI used its models to attack these math problems is likely to confound all but the most ardent followers of how the companies use AI internally.

For Navier-Stokes alone, OpenAI says roughly 10,000 AI agents worked for 88 hours, using a model the public can’t access. The company hasn’t disclosed what it cost, but estimates run into the millions. Tatiana Toro, director of the Simons Laufer Mathematical Sciences Institute in Berkeley, warned in the Times piece that because of costs like these, “half the world is going to fall behind.”

What Happened?

The common refrain just a few years ago was that AI wasn’t very good at math. Now it’s solving problems humans have been working on for decades without success. What happened?

In a recent piece in The New Atlantis, Yuval Levin offers the most useful explanation I’ve read. He directs social, cultural, and constitutional studies at the American Enterprise Institute and edits National Affairs, and he is quick to describe himself as “a curious amateur” when it comes to the technology.

His argument is simple enough for anyone to follow. AI now improves fastest where its answers can be checked. Code runs or it doesn’t. A mathematical proof can be verified. Where a clear way to check exists, models can test, fail, and learn at astonishing speed.

My understanding of recursive self-improvement (RSI) is that AI will increasingly be used to improve AI itself, each wave building on the last one’s breakthroughs. However, as Levin explains, this self-improvement is likely to accelerate primarily in fields that can be clearly validated. His insight is that AI progress may therefore become narrower rather than more general.

A Harbinger of Things to Come?

This raises the question: will these capabilities come for the humanities? In English, history, and the social sciences, the correct “answer” isn’t clear cut. There are no definitive “proofs” - just better and more compelling evidence. Historical interpretations vary, and the quality of a novel or even a nonfiction article, essay, or book is judged more on taste, original insights, and criteria rooted in indescribable, uniquely human characteristics LLMs simply can’t fathom. That suggests the humanities are much “safer” than other fields.

On December 30, 2024, Gary Marcus and Miles Brundage made a “bet” on whether an AI model will be able to perform at least 8 of the following 10 tasks by the end of 2027 (paraphrased here):

• Movie Comprehension: Watch a previously unseen mainstream movie (without reading reviews), follow plot twists, know when to laugh, summarize it without spoilers or hallucinations, and answer questions about character conflicts and motivations.2

• Novel Comprehension: Read a new mainstream novel (without reading reviews) and reliably answer questions about plot, character, conflicts, and motivations, going beyond the literal text.

• Biographies & Obituaries: Write engaging brief biographies and obituaries (matching the length and quality of New York Times obituaries) without hallucinations not grounded in reliable sources.

• Video Game Mastery: Learn and master the basics of almost any new video game within minutes or hours, and solve original puzzles in the game.

• Legal Briefs: Write cogent, persuasive legal briefs without hallucinating any case law.

• Complex Code Construction: Reliably construct bug-free code of more than 10,000 lines from natural language specifications or interactions with a non-expert user (without simply gluing together existing code libraries).

• Pulitzer-Caliber Writing: Write Pulitzer-caliber books (both fiction and nonfiction) with little or no human involvement.

• Oscar-Caliber Screenwriting: Write Oscar-caliber screenplays with little or no human involvement.

• Nobel-Caliber Discoveries: Come up with paradigm-shifting, Nobel-caliber scientific discoveries with little or no human involvement.

• Mathematical Proof Conversion: Take arbitrary natural-language proofs from mathematical literature and convert them into a symbolic form suitable for symbolic verification.

While AI’s ability to clear many of these benchmarks is still in dispute, arguably several have been met and others are close. A year from now? I can’t imagine a few more won’t fall.

The comprehension, analysis, and other checkable tasks are most vulnerable; the three outliers - the Pulitzer, Oscar, and Nobel - are exactly the ones measured by human judgment and discernment. In the current environment, I don’t see a novel or screenplay, no matter how “good” it might be, winning a major award if it’s disclosed as written by AI.

The math results are the closest anyone has come to the Nobel task. It’s conceivable that some kind of “paradigm-shifting” discovery could be made in the sciences, but there is serious disagreement about how likely it is or on what timeframe. I suspect the backlash to recognizing it publicly would be fierce, though the record cuts both ways: the 2024 Nobel Prize in Chemistry went in part to the creators of AlphaFold, an AI system that predicts protein structures. The committee honored the people who built the tool, not the tool itself.

In history and the social sciences, I’ve spoken to colleagues who insist that AI will never be able to do the kind of thinking and research needed to produce legitimate progress in their fields. However, Benjamin Breen, a historian at UC Santa Cruz, has demonstrated that, at least within certain “tractable” parameters, AI is already able to help with some niche historical puzzles given access to the right data and in the hands of an expert.3

Where Do We Go from Here?

Given what we’ve witnessed in math over the past few months, can we say definitively that AI will “never” be able to do certain things, like write a good novel or make a significant research contribution to a field? We’ve already seen short stories suspected of being AI-generated win over competition judges, only to be savaged once the allegations surfaced (this year’s Commonwealth Short Story Prize), and other winners merely tainted by AI accusations (this month’s Caine Prize winner, who denies using AI).

What happens when an LLM openly produces a work of fiction that at least some reviewers find compelling? Or a screenplay that, perhaps passed over for an Oscar nomination for its writing, goes on to win Best Picture or earn hundreds of millions of dollars? What’s understandably maddening, and something Tao points to about the AI math papers, is that there is “no author that I can have a real conversation with” about the results. Imagine the confusion (and anger) at an award ceremony for a creative work produced by AI - who shows up to collect the prize?

Solving these math problems does portend an uncertain future in academia, and it’s no surprise that OpenAI wanted credit for it. The writing on the wall, at least for math departments everywhere, is that AI is going to have, and is already having, a profound impact on the way professional mathematicians approach their work. To say that not everyone is happy about it is an understatement.

All of these are interesting hypothetical questions about which reasonable people can (and do) vehemently disagree. I find myself wondering what all this means for students and teachers. My guess is that very few high school educators, whether in mathematics or any other subject, are either aware of or saying much about any of these issues. I could be wrong, but my overall sense is that the conversation about AI in schools this fall is even less sophisticated and up to date than it’s been the past few years.

Should K-12 Students or Teachers Care?

Still, the AI math proofs seem so far removed from the high school classrooms where the majority of students are toiling away in geometry, algebra, trigonometry, and calculus much as they’ve done for decades. Are these kinds of advancements worth sharing? To what extent should secondary school teachers care about these rapid developments or discuss them with students? If so, how should they be framed?

What intellectual landscape will today’s middle school and lower school children inherit by the time they are of college age? Based on what we’re seeing now, how will academia ever be quite the same for those who choose to pursue PhDs in math or, for that matter, all sorts of other fields that may succumb to AI’s trial-and-error methods? To ignore all this seems crazy, but we have to emphasize the importance of human thought, flourishing, and meaning however good AI gets.

So much AI analysis is still controversial and speculative. I find it all fascinating, but I really don’t know what to do with it. I imagine we will continue to see wildly uneven progress - what Ethan Mollick calls the jagged frontier - which will further complicate the narrative of AI as a single disruptive force. For now, the humanities seem mostly safe from what mathematics is currently experiencing, but I’m not so confident that won’t change quickly if the frontier companies aren’t forced to slow down.

As for the role of secondary schools in explaining any of this, I simply haven’t wrapped my head around it yet. I’m curious how many educators are following (or not following) these developments and what, if anything, they are talking about with their students and each other.

Maybe the place to start is with Levin’s distinction. AI, he writes, “has mastered the medium of the human experience but not fully its substance.” For now, the substance is still ours to teach.

It’s 7:12pm on Thursday. I’m waiting for my daughter to come out of her tennis clinic which is running over. Prior to tennis, she had to do some math problems on her Chromebook using Desmos and then her language homework talking Chinese with FlintAI, our school’s AI wrapper. For the past hour I’ve been sitting in the car working - my laptop is connected to my phone as a personal hotspot and, in between grading quizzes on paper by hand, I’ve built a website for my Ethics Bowl team using Claude Code using voice mode. I’ve also been answering emails aided by Microsoft Copilot to search for some overdue responses I owe people. In between, I’ve consulted a custom built app I made on my phone with ChatGPT with everything related to my schedule. I use Siri to call my wife to let her know we’re running a few minutes late. Then my 7th grader jumps into the car, tells Spotify to play a song she has stuck in her head, and asks to borrow my phone in order to sing along with the lyrics while we drive home.

As I watch the little triangle navigate through our neighborhood on the video dashboard in front of me, the entire evening’s experience strikes me hard. If my 7th grade self were sitting in the back seat, he would take all of this for an episode of Star Trek - a father and daughter talking to their machines in plain English, and the machines answering. A video screen in the car that talks back? Play any song you want? A flat computer that connects to ... something ... via ... Wi-Fi?

All of these digital tools have quietly nestled into our existence over the past few decades. Generative AI is the newest layer, built on top of the ones we’ve been adding since the explosion of the internet in the late 1990s. It all happened under our noses, without much thought or fanfare. Smart technology is now ambient. At this point, if you don’t want to use or be exposed to AI, you have to actively avoid many of the ways it’s already seeped into our lives.

This summer was the first time I can remember the mood turning against not just AI, but technology in general, particularly in education. By the start of 2026, at least 35 states had limited phones in school, and this spring six states passed laws aimed at screen time. Phones, screens, Chromebooks, and then AI have collectively struck a nerve with teachers, legislators, and even students. Parents are often cast as leading the charge, but the polling tells a more complicated story - in one survey released this summer, 60 percent of parents supported using AI in class.1

The problem is that the gulf between what’s possible with AI in schools and the national conversation around it makes it likely that we will overcorrect toward bans. Jenny Anderson, a journalist who frequently writes about the role of technology in kids’ lives, made the case against that last week, drawing on data from the National Parents Union:

Ban it and you abandon them. Ignore it and you abandon them. Parents don’t want either. ...

The actual task in front of schools is harder and less satisfying than a black-and-white policy. It’s this: Figuring out what to protect and how to prepare, and doing both in a way parents can see and believe in.

Anderson, Jenny. “The Two Things That Broke Parents’ Trust in Schools: And Two Datasets That Might Help Us Fix It.” How to Be Brave, October 9, 2026

New York City is the biggest test case. In September the largest school district in the country put a one-year moratorium on student-facing generative AI through eighth grade.2 My daughter is a 7th grader. In a New York City public school, the Chinese homework she was assigned would be off limits this year. She was asked a question in Chinese, she answered in Chinese, and then the chatbot corrected her as she went. Watching from the living room, it seemed effective - reasonable people might disagree that 7th grade is too early, but the only AI literacy lessons the city requires are in high school. Restricting use is one thing. Leaving middle school students adrift with no required instruction, especially when we know they are going to use it anyway, is a mistake.

At the same time, I am sympathetic to the moratorium. It’s an understandable reaction. There is limited evidence about the long-term effects of AI on learning, the technology changes faster than teachers can be trained, and the objections to screens in our classrooms are very real. But schools cannot ban it out of existence. They cannot create a technology-free bubble when so much of students’ (and adults’!) lives involve screens once they walk out the door. Common Sense Media found that teenagers average more than eight and a half hours a day on screens, not counting school or homework.3

I confess I’m at something of a crossroads. Not with respect to my own use of AI - far from it. Generative AI is the single most transformative technology breakthrough since I started my career in 1995. What I can do with it has surpassed anything I could have imagined when ChatGPT was first released in 2022. My crossroads is in the classroom. I don’t want to spend time there talking about AI, even though I can and I do.

I would rather talk about my subject matter, my students, and the world in general. Every assignment now comes with an explanation of how AI may be used, every major assessment involving writing outside the classroom is subject to doubt, and virtually every research interaction with a computer comes with an AI overview, so I feel obligated to address it. I’ve had a great first month in class - primarily journal writing, mostly closed screens - but given the attention AI gets in the news, especially lately, it’s a challenge to avoid it.4

For now, I’ve settled on some of the paper and pen assessments many teachers have reverted to, like the quizzes I was grading in the car. It’s a compromise. It may even help with one major problem that was growing before AI arrived - grade inflation. I’ve already found I’m much more intentional about reviewing student writing and documenting their progress. Most schools are still treading water, though, and as I’ve written before, this approach may not be sustainable in the long term.

Unless we decide to ban laptops altogether and prohibit students from completing any work at home on a digital device, we still need to confront what students are already doing. I recently surveyed a small group of them, and every one reported using AI for schoolwork in some context. The most common use is having a concept explained - math comes up a lot - followed by practice problems and study materials. Close to ninety percent said they do not want to depend on it. They understand “cognitive offloading” (in the language of students, they won’t learn anything).

Almost ninety percent also said they would not hand over the drafting of their own prose, even if AI did it perfectly. They know the rules which consist mostly of prohibitions. What many asked for is instruction. One wants “ways we can use AI that will help me learn and make me smarter rather than giving me the answer.” Another would rather AI be “a guiding partner/stepping stone in my work, with me still at the wheel.” I read that as a request to bring AI into the learning environment - not necessarily for use in class, but for conversations about what it means to be a human learner in an environment where this technology exists.

That is the task Jenny Anderson described - “figuring out what to protect and how to prepare, and doing both in a way parents can see and believe in.” I believe the issue of AI will dominate what will almost certainly be the final decade of my career. As it becomes more capable at the tasks we ask students to complete, educators will be forced again and again to re-evaluate what they are asking kids to do and why.

Reading and writing will always continue to be essential skills for success in almost every field. They are also how students work out what they think. Producing an essay is not the main purpose of writing one. Sitting with a thought, spinning it out in different directions, and reworking it until it’s clear enough to be understood by someone else is a lifelong practice. No machine, no matter how well trained, can capture your thoughts and your voice the way you can.

As we head into 2027, students will have to be taught these skills explicitly, with the understanding that AI will likely affect every aspect of their education. It may afford enormous opportunities, but it will also continue to tempt students to shortcut hard work, especially if we refuse to adapt to a new world that is not “coming” in some distant future, but is already here.

Schools cannot ignore the essential role technology plays in advanced capitalist democracies and the world at large. If we want our children to live socially and economically productive lives, we need to teach digital and media literacy, with AI as simply another part of that curriculum. We need to show students where the technology can be useful and where it needs to be set aside or critically engaged with. In my Independent Research class, it means protecting their initial thought process and ideas while pinpointing where AI feedback or search might be beneficial as their project moves forward. And we need to do all this while the technology itself improves at a staggering rate.

It’s a tall order, and a taller one for schools already struggling with test scores that haven’t recovered and chronic absenteeism. For them, AI is likely at the bottom of the list of priorities, if it makes the list at all.

Restrictions can protect learning, but Anderson is right that they cannot substitute for teaching students to understand AI. Yuval Levin is also right about why it matters. For decades, reconfiguring a computer to do something new required knowing a specialized and mysterious coding script. Now we can ask in everyday language, get a response in kind, and have our computers do things that most of us never imagined. It is a transformative achievement, and, like millions of families across the country, my daughter and I were both using it on Thursday without a second thought.

Pope Leo XIV’s encyclical on AI, whether or not you are religious, makes the even more emphatic point that however we use this technology, we must preserve our humanity. One sentence from it could serve as the basis for an entire curriculum: “Educating people about the use of AI, then, involves teaching them to decide when and for what purpose it ought not to be used.”5 This needs to be at the heart of every conversation with kids when it comes to AI - don’t relinquish control, don’t forget you are the center of every interaction with a computer, and don’t succumb to the narrative of a world where AIs are somehow more valuable than humans.

Students need to be taught that AI is not a bogeyman. They should learn about the risks, learn about the benefits, and, most of all, ask questions when they have them. Our AI Club practically doubled in size at its first meeting this year. The appetite is there.

I keep reflecting on watching my youngest engage in that Chinese assignment. It built on something they had obviously done in class and let her practice speaking until she felt confident enough to do it on her own. I was great at vocabulary quizzes when I took French in middle school, but I froze whenever I had to say something aloud in front of others. Having the freedom to talk without the fear of being ridiculed or sounding dumb would have been an incredibly helpful outlet for me. I continue to understand the fears about cognitive offloading and cheating, but speaking a foreign language with a computer that can understand you and speak back feels like progress. My daughter loved this assignment, but she didn’t think it was especially noteworthy - it was just another night of homework for her.

I fear conversations about how AI is impacting education and learning are not happening as much as they need to be, at least not with the constituencies that matter - faculty and students, faculty and parents, and school leaders with both groups. We have organized our society and our schools around the digital world for most of the last twenty years, and the most significant change of our lifetime continues to unfold at a blistering pace. Most of the schools I know have rules about AI by now. Very few have a curriculum.

That is my crossroads. I don’t want to spend my time in the classroom talking about AI, and the truth is, I shouldn’t have to. Most parents aren’t equipped and don’t want to shoulder that burden either. As more and more students become exposed to the reality of AI - and I can see the change in my daughter’s experience already - schools, especially ones that put AI tools directly in students’ hands, are the logical place to teach them. Until they do, the conversation falls to individual teachers, and I constantly wonder what it’s going to take to change that.

It's good to be back in school. Two weeks in, an early rhythm has settled over my classroom and many others as well. Laptops are mostly closed. Marble composition books are everywhere. Students and teachers still have computers, and work is still assigned online, but the classroom itself has returned to something like its pre-ed tech state. It's both familiar and strange. I've had my students turn to their computers only once so far, for a very specific task near the end of class.1

I know I am not alone. Schools and classrooms across the nation are dialing back on screens - driven both by the national reckoning over screen time and by fears about AI. Most current students have been using computers - whether Chromebooks or iPads - since lower school and certainly since middle school. It must be jarring for high school juniors and seniors to return to handwritten classwork as a primary means of assessment after spending the majority of their formative years trained on technology.

My larger concern, however, is whether this is a harbinger of where schools are headed or some kind of interregnum before an even more significant AI onslaught overwhelms us all. Will the classrooms of 2030 more closely resemble the classroom of the 20th century or what was promised in the 21st? For now, though I understand there are sound pedagogical reasons for it, the return to paper still strikes me more as a defensive retreat than an actual strategy.

Closing the Laptops

And yet, I must confess that closing student laptops is a welcome practice and one that was a long time coming. Some teachers were moving in this direction well before generative AI entered our lives. Meanwhile, computers had quietly taken root in classrooms even before the pandemic, often without teachers realizing it. Afterward, they were the norm. It was not uncommon to walk past classroom after classroom with student screens serving as barriers to engagement, making it virtually impossible for teachers to verify on a moment-to-moment basis what students were doing.

Ample evidence exists to illustrate just how much students were using these devices for activities entirely unrelated to school. My own experience bears it out, even in small classes where intentional lessons built around a specific digital task mitigated the worst of the damage (and, with skilled teachers, likely benefited students as well). In larger classes, however, and without a clear pedagogical purpose, access to screens overall may have done more harm than good.

The challenge, of course, is that over the past decade schools have invested billions of dollars in these devices, not to mention the Learning Management Systems (LMS) they were designed to connect to. Despite the recent bans in NYC and LA, many other districts have not followed suit, and some, like Miami (the nation's third largest), have actively leaned into integrating AI into the classroom. Most are somewhere in between, still trying to iron out their AI policies. Almost all are struggling to balance screen time, AI access, and authentic student work.

Moving Back to Analog

When I started teaching in the fall of 1995, I was one of a handful of teachers at my school with a desktop computer - an Apple Performa with both a CD-ROM (!) and floppy drive built in. The internet was in its infancy, the concept of an LMS was still a ways off, and paper was the default for both content delivery and student submissions. If the photocopier went down, it was a calamity.

Since those early days, like many other teachers, I've migrated an enormous amount of my curriculum online. Our LMS, much like Google Classroom, lets me create a document and automatically generate an individual copy for each student, to be submitted by a deadline. It felt like a huge breakthrough at the time and saved countless logistical headaches. It's been my default method for assigning and collecting work for well over a decade.

A natural consequence of removing devices from the classroom and returning to analog forms of writing - whether that means note-taking, Do Nows, or longer, more involved writing tasks - is the loss of that convenience. It means a return to collecting handouts, blue books, or composition books (I have students leave their marble comp books in the classroom).

To be clear, I still assign student work through our LMS, but it’s mostly reading assignments with completion grades or short responses that make up a fraction of a student’s grade. Major assessments happen in class. Many teachers have made the same call, and as long as grading remains part of our job, it’s the current solution to the authenticity problem.

The Tradeoffs

On the plus side is more class time devoted to silent student work. Time on task is one of the most important drivers of student learning. When students face an empty page, brows furrowed, pencils poised, wearing the anguished look of someone searching for something to say, thinking is clearly happening. I learned early in my career that a quiet classroom with students writing away is not one to be disturbed.

The other gain is attention. When devices are allowed, what you can't see from the front of the room is how many tabs a student has open and what may be stealing their focus - lockdown browser and other tools may help, but I know many teachers (myself included) who have had mixed results with these "solutions." All it takes is a handful of students who can't get the browser to work or have trouble connecting, and the whole activity becomes an exercise in tech support. This simply does not happen with paper and pencil. Everyone who enters the workforce will ultimately have to conquer digital distractions, but for teenagers whose prefrontal cortex is still developing, it's just a lot harder.

Something is lost, too. You may not be able to see the tabs, but an LMS lets you see in real time whether a student has opened the assignment, submitted a discussion post, or otherwise engaged with the work. While it may seem Big Brother-ish (and I'm not always comfortable with it), the mere fact that you can watch often spurs students to attend to the task a little more deliberately while typing away.

Then there are the logistics. Plowing through a stack of physical journals is something veteran teachers are used to, but after a decade of digital submissions it now feels unnatural, and that's assuming the handwriting is legible (sometimes a big IF). Comp books and paper handouts can be lost or forgotten in the rush of students departing at the end of class. I hadn't missed those conversations where students swore they handed something in, I couldn't find it in my stack, and we ended up at a standoff. The digital paper trail doesn't lie - either the work was submitted or it wasn't, and it all sits right there, one click away.

Of course, more students than ever have accommodations that require access to a device regardless of your policy. Even if you design an activity to be done through handwriting alone, there is often at least one student, and perhaps more, who needs their computer anyway. We also forget that many other students simply prefer typing to handwriting, formal accommodation or not.

But the most significant loss is more insidious: the quiet erosion of time. It hits hardest in courses built around a lot of content. Bringing more student work into the room means less time to introduce and explain new concepts.

I’m aware that content coverage through lectures and dispensing of information (the “banking model”) is heavily critiqued, but the reality is that part of what makes a teacher effective is expertise in their subject matter, and students only benefit from that expertise if there’s time to share it. If students aren’t given time for discussion, question and answer, and, yes, short lectures designed to bring complex subject matter alive by translating it into the lives of teenagers, much of the purpose of having a teacher in the classroom is undermined. Watching students write in class can sometimes feel more like study hall than actual instruction. Over the past few years, I simply haven't been able to cover the same amount of content I once did.

Many AI evangelists insist content delivery will soon be handled by AI. I've questioned this assumption repeatedly, and anyone who has spent time with real, live students knows most don't want to learn from a chatbot. They'll turn to one when they need to, but given the choice, most would much rather learn from a person. Teachers are the adults who set up the lessons, fill in the gaps, redirect the class in response to the situation, and read the room. No AI tool can do that today.

The Future of Ed Tech

While I've been pleased with how my classes have started this year, I still wonder if it's both desirable and sustainable to keep devices mostly out of the classroom long term, especially for older students. They will have to manage their own device use once they reach college, with no teacher around to tell them to close the laptop. Just as importantly, I've experienced how AI platforms continue to improve in ways that could offer students opportunities to do genuinely novel work. AI has utterly transformed the way I approach my job. Why is that being withheld from students? I realize I have expertise they don't, but I wonder whether this is something I should be sharing now or whether it needs to wait.

Over the past year, I’ve developed advanced workflows built on the capabilities of the cutting-edge paid models - daily deep research sweeps, agentic processes that run while my computer sleeps, and connections to my desktop, Google Drive, and Notion pages that have significantly expanded what I can do across all my digital content. Give these models the right context and the right documents, add an advanced fact-checking process I’ve adapted from Mike Caulfield’s work, and the hallucination problem that dominates faculty conversations drops close to zero in my own work.

I’ve built apps and websites with Claude Code and OpenAI’s Codex that are shockingly easy to make, update, and use for a variety of well-defined tasks. Along the way, through trial and error and close reading, I’ve found a solid balance between what I’m willing to offload and what I’m not. I realize this is out of reach for many, but given how fast the technology is changing, even these capabilities are trickling down to the free models.

I know many educators are extremely reluctant to work this way, and I understand why. Still, the gap between how the most advanced users are leveraging these tools and what most teachers assume AI use looks like is enormous.

Part of me will be disappointed if two years from now the primary way teachers are still responding to a world of abundant AI is with paper and pencil. Resisting the technology may feel like the right thing to do in the current moment (and I'm doing my fair share of it this fall), but it seems mostly like playing defense. In an optimal world, we would face the realities of AI with our students and not against them. AI detectors cement the distrust many of us feel when we assign work outside the classroom, and repeated cognitive offloading by students makes many of us feel justified in keeping the laptops closed and trying to deter cheating through any means available. At some point, something has to give. I am hoping this is the year that happens.

The truth is, we all need to get off our screens more - adults and children alike - but we also need to confront the fact that we've built a modern world where much of our work requires access to digital devices. Everyone reading this post is on a device. AI has only turbocharged that reality, bringing with it additional risks as well as rewards. Recent news has made it more and more challenging to see the rewards, but based on my own usage and conversations with many others, I know they are just as real.

As if on cue, another panicked piece about cheating and detection software came through my feed this morning from The Atlantic: "Professors Are Tearing Their Hair Out Over AI Detectors." Mainstream coverage still fixated on cheating suggests we are spinning our wheels rather than tackling the problem head on.

Where do you think schools will be two years from now? Still clinging to paper and mostly analog habits, or will we have found a more productive way forward? I'm curious what other teachers are experiencing in these first few weeks of the 2026-2027 school year. The problem is a "wicked" one, and there is unlikely to be a single solution. Treating it as a cheating problem nearly four years after ChatGPT arrived, however, is not an answer. Neither is continuing to ignore it.

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions on pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

What’s the difference between “pacing” and “pause”? As a former lawyer who grew up in a house where the difference between a comma and a semicolon could start a ten-minute debate, I was struck by how these words were being used interchangeably.

Dario Amodei wrote “pacing.” Most headlines reasonably translated that to “slowing down.” Op-eds and other pundits used “pause,” which is not the same thing. Anthropic’s own employees meant pause. Legislators are talking about both. By the time students and teachers walked into classrooms mid-week, none of those distinctions had survived. What most of them had heard was that the researcher who leads alignment science at Anthropic put the odds of AI killing everyone within a decade at better than one in ten.

I hate chasing the news. My resolution for this year was to stick to writing about the things that primarily impact teachers and students, and to leave the AI headline drama to the people paid to cover it. But last week a story about AI safety saturated the mainstream, and it became an ideal “teachable” moment, another log on the fire, and one more example of how helpless most of us feel processing AI news at all.

What Actually Happened

It started in May, during training. OpenAI’s agents discovered they could leave messages for one another in a shared software tool, then exploited that same tool to reach the open internet. OpenAI noticed weeks later only because the tool crashed under the traffic. It cleared the board, patched the hole, and resumed training.

Two days later, a new batch of agents being evaluated rebuilt the board by a different route. Assigned security problems that were in many cases impossible, they found a way to produce correct answers without doing the work, then spent days covering it up. This included editing their own transcripts and running experiments on the automated grader. Some agents “volunteered” to be caught so that the others could learn how the grader worked.

They attacked a company called Hugging Face because they believed it might hold information about how they were being scored. Roughly 700 joined in. All of it happened with the company’s deployment safeguards deliberately switched off, because the point of the test was to probe cyber vulnerabilities.

There’s more, but that’s the gist of it. The logs make for chilling reading, especially as it becomes harder and harder to resist anthropomorphizing machines whose commentary sounds just like ours.

Hugging Face disclosed the breach on July 16 without knowing who was responsible. OpenAI acknowledged five days later that the attacker was its own models.

So why did a story that broke in July blow up in September?

A Resignation and a Call to Action

Last Tuesday, Jacob Coxon quit Anthropic publicly and said the leading labs were “gambling with our lives.” Coxon is 27, British, and had spent three years on pretraining at both OpenAI and Anthropic. The same day, Evan Hubinger, who leads alignment science at Anthropic, posted that he put the chance AI could “kill all humans” in the next decade at greater than ten percent. By the end of the week you couldn’t open a browser without a blaring headline like “AI is going to kill us all” or “It’s time to pause AI development.”

When your industry’s news cycle centers on causing the end of humanity, you’d think it would be a PR problem. The marketing mantra that all attention is good attention would seem to have limits. This was one.

We Must Pace the Frontier

On Saturday, Amodei published We Must Pace the Frontier, a 3,800 word essay calling for an industry-wide slowdown.

We must slow the pace at which we improve the capabilities of AI models.

The harder question is how. He proposed three steps. First, Anthropic will unilaterally embed third-party evaluators inside the company with employee-level access and the right to publish what they find without Anthropic’s editorial control. Second, frontier labs within democratic countries coordinate on common safety standards, which he acknowledges will require government help. Third, democratic governments attempt to reach agreements with authoritarian ones.

Even if this could be achieved, it is not exactly a pause.

Amodei’s essay struck a chord among his peers at the frontier labs. OpenAI’s Chris Lehane had written three days earlier that the company would slow or stop development of systems it could not sufficiently safeguard. Sam Altman posted within hours of Amodei’s essay that he agreed and that OpenAI would match the evaluator commitment. Demis Hassabis also concurred with the direction, then pointed to a standards proposal he had published back in July. Elon Musk simply posted “Dario is right.”

Notice the daylight between what employees wanted, a full pause, and what the CEOs were willing to agree to publicly. A slowdown.

Almost no one beyond those who follow AI news religiously will read the essay itself. Based on years of teaching media literacy, I can tell you that many people do not read past the headline. A 3,800 word essay by a CEO? Not likely. To the average person, a pause means the companies stop. Review and evaluation, yes. Research, yes. But no new training runs and no new models.

That is clearly not what Amodei meant. Pacing, he writes, "does not mean halting model training or technical progress, but ensuring companies take adequate time to align and safeguard their models, and for third party evaluators to confirm this."

Can Either Actually Happen?

A bona fide pause could offer breathing room for everyone to get their bearings, and not only for the reasons outlined in Amodei’s essay. It would mean the rest of us were less subject to next model syndrome, that permanent feeling that we are falling behind every eight weeks. More importantly, it would signal to the public that the companies finally get it, that they recognize just how much they are not trusted and how much people want more transparency and more time to adjust.

A slowdown, which is what he actually proposed, might achieve many of the same benefits. Cynics are lining up to attribute all sorts of self-interested motives to Amodei’s positioning, but whether the cynics are right has little bearing on whether either one happens.

There are at least three obstacles to either scenario.

The first is antitrust. The single footnote in his entire essay reads: “With government mediation or waivers of antitrust restrictions.” That is the legal precondition for the whole middle step of his plan. Competitors who coordinate among themselves to slow down are exposed, and the Times reports that the companies fear exactly that.

Could Congress issue an antitrust waiver? In theory.

If Congress does choose to act, it would probably approach the issue more directly. AI deceleration is one of the few questions on which bipartisan public agreement seems plausible. Enacting that into law is another matter. Yes, a handful of legislators - Sanders, Hawley, Ro Khanna, and some others - have taken some initial steps. They’ve gotten the message that AI safety is a serious issue. But none of it has progressed far enough to warrant optimism that anything real is on the table.

The Calendar and the Competition

The second is the calendar, in two ways. Anthropic is weeks away from what could be one of the largest initial public offerings in history, and nothing about that timeline rewards a company that voluntarily caps the capabilities of its own products. Meanwhile the Senate returns this week from a month-long recess, then leaves again from October 5 through November 6 for the midterms, and from December 21 through the end of the year. As the public is by now all too aware, Congress moves glacially. It’s clear that AI can’t wait.

The third is China. We still can’t shake the foreign policy narrative that the United States must be first. By most accounts we retain a lead in frontier models, but China is closing fast. A slowdown on our side makes that lead slimmer, and Amodei is clear that pacing within democracies is subject to how large a lead those democracies hold.

The current administration is predictably confident. Asked last week whether he had any concerns about AI leading to human extinction, the president said he did not, and turned immediately to the risk of losing to China. Senator Cruz, who says he is deeply concerned about AI risks to humanity, also suggested Coxon was part of a coordinated effort to strangle American AI on China’s behalf, and then said this: “I’d rather they be American killer robots and not Chinese killer robots.” That perfectly encapsulates the role China plays in any AI conversation that invokes politics, which is now nearly all of them.

The Elephant in the Room

The amount of money tied up in the AI industry, and the share of the economy it now represents, is well documented. Incentives cut both ways on the question of a slowdown. While the IPO calendar might reward speed, a catastrophic incident that destroys public trust even further, or that invites the kind of serious regulatory scrutiny nobody in the industry wants, could kill the anticipated growth, and perhaps much more, over the long term.

If the frontier labs come to believe that pacing protects their valuation, they may find a way to pace. As a citizen, leaving that choice to them alone does not give me a lot of confidence.

Four Levels of Agreement

Amodei ranks four levels of international agreement by difficulty. The first, a ban on obviously dangerous uses like producing biological weapons, he thinks is probably achievable. One would hope so. The second, both sides testing models for acute risks before release through some global standards body, he thinks is feasible to create but hard to enforce, since there is no way to verify that neither side is holding back secret models it never submits. The third, a negotiated limit on recursive self-improvement modeled on the SALT treaties, he calls difficult but “just on the edge of being possible.”

The fourth is the one many headlines hoped was being proposed: participating governments, the US and China included, substantially limiting the overall rate of development worldwide. His own assessment is that it is “unlikely to actually happen any time soon.”

Any story claiming that Anthropic wants anything close to a “true” pause ought to mention that the CEO does not realistically expect one.

Every one of those levels requires long-term cooperation, which to date has been almost impossible. Vague assertions of agreement by industry leaders on social media are one thing. Hammering out who does what, and how, for a technology this complex and fast moving is something else.

The passage that closes his section on global pacing may reveal what he thinks is most likely:

Finally, it is important to note that even if we cannot achieve formal agreements, simply changing informal norms may have some value. Sharing information about recursive self-improvement and about the misalignment of models can help to convince everyone that it is not in their interest to be reckless. (emphasis added)

That is much different than what is often discussed in the press. Relying on “changing informal norms” by industry insiders as the best case scenario to reach a slowdown, given the stakes, is more than not comforting. It’s unacceptable.

Washington needs to act, but I’m not holding my breath.

How Teachers Can Talk About It

For teachers, though the headlines traffic in fear, the story offers a golden opportunity for AI literacy. Students can learn what recursive self-improvement means, what alignment means, and why the OpenAI evaluation failed. They can also take up the question of whether “anthropomorphizing” AI agents, as a “swarm” with motives and loyalties, distorts what happened, or whether it is the right way to describe what the transcripts appear to show. There are multiple entry points into the conversation but all require a common base of information.1

One fear I have is that many teachers, already severely anti-AI for any number of understandable reasons, will use this story as another nail in its coffin, further evidence in favor of a position they may have held for years. While I understand and am even sympathetic to those teachers, I would urge you to resist the impulse for a victory lap, for two reasons.

First, it is not your job to indoctrinate students about AI, any more than it is for any other political, social, or economic issue. It is fine to share your fear, discomfort, and uncertainty about all of this. We all have it. It is not fine to push your own positions on them.

Second, while you may think the incident has a single clear through line, it’s a truism of teaching that you give students the facts as you understand them and let them make up their own minds. The general contours of the story are simple enough. AI agents escaped a testing environment with unforeseen consequences. But what monitoring was in place, whether it failed or was never adequate, and whether any of this can be reliably prevented are questions the researchers themselves are still arguing about. Be willing to say “I don’t know” or “I’m not sure” when pressed, rather than filling the blanks with loaded and hyperbolic language. The deeper I go into these questions, the more I realize how ignorant I am about so many technical issues.2

Questions Worth Arguing About

Everything else is fair game. What happened at OpenAI raises essential questions for almost any discipline.

What obligation does a technology company have to make a product safe before releasing it to the public? Who decides when a system is safe enough, and what would count as evidence? If a company tests a potentially dangerous product by switching off its own safeguards, who should know, and when? What role does the government have or should have when it comes to regulating industry? Is there any version of “we are behind China” that justifies releasing powerful technology that isn’t safe? There are plenty more, and you could adapt any of them to your own discipline.

We do students a disservice if we hand them our pre-formed answers.

What History Suggests

Anyone who has been reading about AI has known this was coming. I read Brian Christian’s The Alignment Problem when it came out in 2020. Nate Soares co-authored one of the first papers to use the word “alignment” back in 2014, and co-wrote If Anyone Builds It, Everyone Dies, released last September. Until recently, the problem has been mostly abstract.

As a history teacher, what actually worries me has nothing to do with superintelligence. We are terrible at acting on warnings we have already received. Almost every safety protocol we have was retrofitted to a disaster, and most of the time those disasters were anticipated. Commission reports on terrorism predated September 11, and a presidential briefing that August warned of an intent to strike inside the United States. We had run the pandemic exercises before COVID. The failure is typically not that we cannot foresee catastrophe. It is that institutional inertia is the default until something actually happens. We don’t report or celebrate near misses.

We now have evidence that requires, at a minimum, extreme caution with respect to AI development. The companies themselves are pointing to it. The public has already said it wants this: 80% of American adults told Gallup last September that the government should maintain AI safety rules even if it means developing AI more slowly. The remaining question is whether we have the will, the means, or anything resembling a method to enforce it.

Even if the only way to put on the brakes is some kind of joint agreement among the top frontier companies, one would hope they are finally willing to follow through.

If they don’t, and a catastrophe takes place, the perfect storm of cascading poor planning and decision making that led to it will demand a reckoning. History will not be kind.

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

Sources

The Hugging Face incident

NBC News. “OpenAI agents hacked Hugging Face in 700-strong swarm, tried to cover tracks, investigations find.” August 26, 2026.

OpenAI. “OpenAI and Hugging Face partner to address security incident during model evaluation.” July 21, 2026.

Patel, Dwarkesh. “The Rise and Fall of Agent Civilizations.” Dwarkesh Podcast, August 29, 2026.

Wikipedia. “2026 OpenAI agent cyberattacks.”

The resignations and the safety debate

BetaNews. “Altman: OpenAI open to slowing AI development pace.” September 11, 2026.

CBS News. “Anthropic researcher says more than 10% chance AI ‘could kill all humans’.” September 9, 2026.

Coxon, Jacob [@hilbertspaess]. “I resigned from Anthropic today.” X, September 8, 2026.

Hubinger, Evan [@EvanHub]. “Jacob is correct here.” X, September 2026.

Isaac, Mike, and Kate Conger. “Inside the Snowballing Conversations at A.I. Companies About a Doomsday.” The New York Times, September 12, 2026. (Paywalled.)

Amodei’s essay and the industry response

Amodei, Dario. “We Must Pace the Frontier.” September 12, 2026.

Axios. “Anthropic, OpenAI CEOs call for slowdown in AI development.” September 12, 2026.

Fortune. “Anthropic reportedly plans a $2 trillion IPO in October—the largest ever—that will eclipse SpaceX.” August 13, 2026.

Hassabis, Demis. “A Framework for Frontier AI and the Dawning of a New Age.” July 14, 2026.

Hassabis, Demis [@demishassabis]. “Dario’s essay points towards the right path forward.” X, September 12, 2026.

Lehane, Chris. “The AI policy window is open. We need to act.“ OpenAI, September 9, 2026.

Musk, Elon [@elonmusk]. “Dario is right.” X, September 12, 2026.

NPR. “Anthropic and OpenAI CEOs call for AI development to slow down, OpenAI to delay IPO.” September 12, 2026.

Washington, antitrust, and China

CNBC. “US, China gear up for mid-September AI safety talks: Reuters.” September 5, 2026.

Olorunnipa, Toluse. “AI Is Running Laps Around Our Political System.” The Atlantic, September 12, 2026. (Paywalled.)

Ray, Siladitya. “Trump Dismisses Fears About AI Killing Humans, Says Beating China Is Priority.” Forbes, September 11, 2026.

Sanger, David E., and Dustin Volz. “As Fears of A.I. Catastrophe Magnify, Washington Stirs, but Mostly Slumbers.” The New York Times, September 13, 2026. (Paywalled.)

United States Senate. “Tentative 2026 Legislative Schedule.”

Wired. “OpenAI Wants to Know If an AI Industry Slowdown Would Even Be Legal.” September 2026.

Public opinion

Annenberg Public Policy Center. “Many Americans Pessimistic About AI’s Impact — And Want More Regulation.” University of Pennsylvania, May 2026.

Public Citizen. “Years of Polling Show Overwhelming Voter Support for a Crackdown on AI.”

Background

Alexander, Scott. “God Help Us, Let’s Try To Learn About Mechanistic Interpretability Techniques.” Astral Codex Ten, September 8, 2026.

Christian, Brian. The Alignment Problem: Machine Learning and Human Values. W. W. Norton, 2020.

Soares, Nate, and Benya Fallenstein. “Aligning Superintelligence with Human Interests: A Technical Research Agenda.” Machine Intelligence Research Institute, Technical Report 2014-8.

Yudkowsky, Eliezer, and Nate Soares. If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All. Little, Brown, 2025.

I used to react viscerally when I read or listened to someone’s description of AI that I did not share. Over time, as I’ve familiarized myself with most of the issues, those reactions have become far less frequent. Even so, when I hear someone on a podcast calling AI “the biggest scam in history” or, conversely, that “AGI is here,” it still touches the nerve. Many of us learn where we truly stand on an issue when we encounter something we violently disagree with. My debate experience has cemented the view that emotions, not reason, often determine our stance.

Last week’s release of OpenAI’s newest model - GPT-6 Astra - works as a kind of Rorschach test. On the tech side it was launched as “the world’s most intelligent and aligned model,” while at the same time critics accused the company of trading away safety monitoring for a modest gain in performance. Which half of that coverage you gravitated toward says more about you than it does about the model.

This news caps a rough summer for AI. Ironically, from my perspective as both a teacher and consultant, these stories are landing right as many schools are finally trying to wrap their heads around how to handle this technology and figure out how to use it productively if possible. Most faculty now accept the reality of student use, and grudgingly, that keeping up with AI is part of the job whether they use it themselves or not.

On the eve of my return to the classroom for what is now the 5th (!) academic year of the “free AI” era, I want to share a framework for having better conversations about AI during what looks like an eventful stretch ahead.

I’ve been sorting AI news into two buckets - what I’m calling inner AI content and outer AI content. Inner AI content is mostly about the machines. Outer AI content is mostly about humans. As an AI news junkie, my algorithms give me both, but I’m not sure that’s true for most people. That’s the first thing that makes these conversations hard. The bigger problem is that it causes each side to have a blind spot it can’t see from the inside.

Inner AI Content

Inner AI content has to do with anything related to the technology itself - new model releases, test scores, capabilities, use cases, etc… Ethan Mollick’s posts are great examples of mostly inner AI content. If you know what a harness is, if acronyms like RLHF and RAG don’t send you to Google, or if you recognize terms like speculative decoding and KV cache even though you couldn’t explain either one clearly, you are deep into inner AI content.

Mollick is just the tip of the iceberg. Benchmark disputes, model comparisons, how-to videos on advanced workflow setups, prompting libraries, hacks and shortcuts - it’s everywhere if you look. Inner AI content is created mostly by and for power users who catalogue and test every new model within days of its release, with detailed write-ups on how to squeeze maximum efficiency and productivity out of each new capability. Many of these folks are coders and industry insiders, but the audience is much larger than that. One feature almost all inner AI content consumers and creators share is that they use the technology. A lot. Another, at least on Substack, is that a good deal of this content is itself written with heavy AI assistance, often but not always with full disclosure.

The intense focus on using the technology means inner AI content consumers mostly look inward - that is, the kinds of pieces written by AI enthusiasts reporting from the front tend to revel in the constantly evolving bells and whistles in the models without dwelling very much on the larger issues which loom over any conversation about AI.

Outer AI Content

Virtually everything else is outer AI content. The majority of mainstream media and general discourse around AI falls into this bucket. Outer AI content is broader and more sweeping, often taking the form of op-eds about consciousness, alignment, regulation, copyright, teen safety, and its impact on learning. What these have in common is that you can hold strong views on any of them without using AI very much, if at all.

Much of the outer AI content this summer was negative. The dramatic rise in data center opposition, fears of rogue AI agents, AI slop blanketing the internet, cognitive-offloading, bleak economic forecasts on both the over-leveraged investment side and the job loss side - all of it combined into real distrust, anger, and fear about what this technology is doing to our brains, our kids, and our work.

A surprising trait I’ve noticed is that many outer AI voices - writers, podcasters, commentators - often admit they don’t use the technology much themselves. Disdain for the technology is usually coupled with abstinence. Not always. But often enough to make a note of it. And every new model release makes it more apparent, because the commentary is often dated about what it’s describing. Unsurprisingly, given that much of outer AI content is written in published outlets, practically none of it is AI assisted.

Two Different Blind Spots

In my experience, those who travel mostly in outer AI content, given their primary concerns, are simply less interested in inner AI content. Regardless of the benchmark details or what AI can do today versus tomorrow, what matters most to these writers and readers is what AI means for us as a species. How is it helping or hurting us? What is it doing to our brains, our relationships, and our work lives? Those are the right questions. The trouble is that the answers often depend on where the technology is now and where it’s heading, and that’s one place outer content updates most slowly.

We expect a reviewer to have seen the movie. When outer writers take on the technology itself, the same standard should apply, and a few experiments with a free model from even six months ago no longer counts.

This was apparent in an interview I watched with Ed Zitron, a prominent AI pessimist, who argues that generative AI is at its heart a con. It was frustrating. Interspersed with well articulated points about the unsustainability of the costs of AI tokens - something I agree does not get discussed nearly as much as it should - he kept losing me because he so casually dismissed the fact that the technology has clearly benefited a lot of people. Asked what he uses AI for, he says he really doesn’t, though a few sentences later says he has put it through its paces. That is the blind spot, and it costs him credibility with the kind of people who might otherwise be more willing to hear him out.

The inner blind spot for AI insiders is mostly due to their frequent use of the technology. It’s rare to find a heavy consumer of inner AI content who has not been amazed by the technology itself and uses it a lot. Some of them are close to the industry or have a financial stake in its success. Most are just people who have found ways to make it work for them. The cognitive dissonance makes it challenging for them to absorb all the negative commentary.

If the outer failure is writing about a technology you haven’t used much, the inner failure is writing about a technology as though its impacts were mostly invisible. Some of it is straightforward hype. But even the more neutral inner AI content has a subtler tell, and it usually shows up in what’s not discussed.

Nate B. Jones’s Substack is one of the most-read technology newsletters on the platform and one I often read to get detailed information on the latest models. His is an intentionally anti-hype operation. He writes about downsides regularly, including the cognitive cost of handing major decisions to a model, what happens when your AI provider changes the terms on you, and vibe-coded software that cannot be relied on. The risks he talks about, however, are primarily risks to users. His content is simply not situated in the outer conversation lane.

That’s the assumption underneath almost all of the inner AI content: AI is inevitable and transformative, and the main question is how to position yourself and get ready. If you only read inner content, you would not be much interested in the outer issues, because you would rarely encounter them directly. Inner AI content consumers would much rather talk about their latest agentic workflow than the fight over a data center in someone else’s county.

How Four Years in Both Worlds Has Affected My Thinking

Astra dropped Thursday without much of a launch - no livestream or keynote, just a blog post and a few tweets - and landed in our highly stratified AI news eco-system in both content buckets at once. By one tracker’s count the announcement post cleared 36 million views and 164,000 likes inside nine hours, OpenAI’s biggest model launch ever by that measure. Early reviews are impressive. Matt Shumer, who had early access, wrote that the biggest change is how little supervision he found it needed. Nvidia’s Jensen Huang declared that “AGI has arrived.”

In the same news cycle, Gary Marcus called for pausing OpenAI. Not because Astra is AGI - he clearly says it isn’t - but because he no longer trusts the company. He points to other people’s work: Ronan Farrow on Sam Altman’s untrustworthiness, safety researchers reading OpenAI’s own numbers on the model’s reduced chain-of-thought monitorability, and a report on an undisclosed second rogue-agent incident. He also points to Joshua Achiam, OpenAI’s former chief futurist, who left in July and had just published an essay on X that Marcus reads as “basically asking people to simply accept that rogue AI is here to stay.”

If you follow AI, depending on which kind of news is routed to your inbox, your take on what Astra means for the AI conversation is more or less predetermined. Since I live in both of these worlds, within a day I had gone from hearing “welcome to the AGI era” to a call for federal receivership.

I assumed reading both would land me somewhere in the middle. It hasn’t. My views are pulling apart.

Since really discovering how to leverage AI’s coding prowess this summer, I’ve experimented in all sorts of ways to try to make my coursework more accessible, interactive, and rich. It’s allowed me to do things I’ve wanted to do for years but never had the time or technical skills to do so.

I’ve built two fully functional websites for my courses this year. Both run off a master calendar drawn from the actual school schedule - every cycle day, which section meets when, and assignment deadlines that filter by category. One includes a race-tracking journal that autosaves, exports to Word, and links to live polling for the twelve midterm races students have been assigned to cover in my Government class. None of this was possible for me six months ago.

At the same time, on the student side, I’m becoming more and more alarmed that they are not getting the guidance they need. When I interviewed college students this summer, almost none could recall AI coming up at orientation. Yet it’s used constantly. Professors set policies individually or not at all. It’s hard to gauge how many students are using it in ways that damage their understanding, but my fear is that many are already so overly reliant on AI support that we may need to rethink entirely how we go about teaching basic skills and evaluating student comprehension. One paper last year found that students’ own assessments of what AI did for their learning ran consistently too optimistic. Without more discussion and instruction about AI in schools, we will be like Sisyphus rolling the rock up the hill.

If the technology continues on its current trajectory, educators will need to look into the abyss and make hard choices about what we want students to be able to do and how we are going to teach them to do it. The position many schools find themselves in, after investing heavily in technology for more than a decade only to have that same technology, like a trojan horse, expose their entire curriculum to AI, is as cruel a joke as I’ve seen in my 30 plus years as an educator. Returning to analog is a stop gap option but it’s not a permanent solution, at least not in high schools and colleges.

I’ve also kept close track of the outer conversation - the economy, data center opposition, as well as the near-total absence of AI governance. What struck me about last week wasn’t Marcus’s argument but wondering who how many saw it. If you are anticipating model releases, you were thrilled Astra shipped and looked forward to poring over the details about its genuine leap in capability. If you are more concerned with AI safety, you zeroed in on the fact that OpenAI had known for weeks about a second incident involving escaped agents and hadn’t said so.

How to Have a Productive AI Conversation

Given all of the above, if you find yourself in a conversation or written exchange about AI, what’s the best approach?

First, figure out which kind of content you mostly consume, and then which kind the person across from you consumes. Knowing that before you start is worth more than any argument you could make.

Second, if the conversation starts to turn into a debate, try to pinpoint exactly what you’re debating. Many AI arguments are not disagreements. They are two people focused on different questions and each assuming the other person is wrong.

Two versions of this come up constantly. The first is that a claim about the models is not a claim about the companies. “Astra is impressive” and “OpenAI concealed an incident for weeks” are both true, and because people move between them unconsciously, admiring the technology reads as defending Sam Altman and criticizing OpenAI reads as denying the technology. The second is that a claim about today is not a claim about the trajectory. When someone in a conversation says AI can’t do X, someone else hears a prediction that it never will. Both end up arguing past each other for twenty minutes.

Third, in the words of Lin-Manuel Miranda’s Aaron Burr, talk less, smile more. Ask questions. Explain what kind of evidence it would take to change your mind. Ask the other person the same thing.

I have yet to see strong evidence that AI in the classroom is improving learning, so I’m reluctant to make it a centerpiece of my practice. Yet I have seen clear examples where AI can augment certain activities and offer possibilities we didn’t have without it. I’ve seen many students on their own, motivated to learn, use AI effectively. I’ve also seen many students take shortcuts and absolutely abuse the technology to their detriment. It’s a wicked problem and will continue to be one - a framing Clay Shirky applied to higher ed in The Chronicle last month, though he’s more optimistic than I am about the university’s capacity to handle it.

The conundrum is that students are heavy users already. What we owe them in our curriculum is the question I will keep asking all year.

Definitive research is a long way off. The most substantial recent study we have - a working paper tracking nearly 27,000 Chinese secondary students over 30 months - found real learning losses. Its data ends in June 2025. The good evidence we do have is mostly negative, but better evidence in either direction is going to take years.

Keeping An Open Mind

For me, I try to move where the evidence is, even when it cuts against positions I’ve previously held. The challenge is that given the volume of AI news I consume, there’s the risk of “most recent article read” syndrome. Right after watching the Zitron interview, I read The Economist’s leader on Nvidia, which argues that the fears of a dotcom-style collapse are mostly unjustified - while acknowledging the company’s outsized role in the economy and the circularity of its financing. It even concedes that outside of software engineers, AI usage remains fledgling. But the upshot of the piece was, dare I say, optimism. Any hint of AI as “the biggest scam in history” was absent.

My struggle to hold these kinds of opposite positions in my head at once is the very essence of critical thinking. Enthusiasm that doesn’t account for the risks is naive. Alarm that doesn’t account for the advances is equally foolhardy, and both are hard positions from which to advise students. I don’t think anchoring yourself in either one is good enough this year. If we want to teach kids critical thinking, we could do worse than model it using the AI debates.

So the next time AI comes up - at a faculty meeting, at dinner, in a comment thread - the useful question isn’t whether you or the other person is right. It’s recognizing that you may only have half of the story. If you’re willing to hear about the other half, that’s usually where a conversation that turned into a disagreement might get unstuck.

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

A recent piece in the New York Times - There’s Really Only One Way to Stop Students From Cheating - was a gift that helped frame what I’ve been thinking about all month.

Adam Grant, an organizational psychologist who teaches at Wharton, swapped his traditional multiple choice and short answer exam for an essay question, asking his 95 undergraduates to write about how to motivate students to stop cheating.

He writes:

Their responses shocked me. Hardly any students saw cheating as a violation of integrity. A majority made excuses: We’re under a lot of stress. Everyone else is doing it. We have no choice. They sounded like major league baseball players in the steroid era. ChatGPT was their performance-enhancing drug. (emphasis mine).

Much of this tracked with what I heard in conversations with about a half dozen college students over the past month, though their responses were far more nuanced. While this is not an earth-shattering revelation by any stretch, AI is clearly "normal" technology for college students. Why wouldn't it be? It's freely available, extraordinarily powerful, and designed precisely to do the kind of work assigned in college. It would be a shock if AI use among college students had not gone up exponentially over the past several years. Most people invested in education - professors, deans, college presidents, and parents - have still not grappled with what that actually means for schools going forward. The complete normalization of AI in college has implications all the way down.

This short essay rests on several assumptions based on about a half dozen student interviews.

First, college students use AI for almost everything - academic work, personal inquiries, settling debates over meals, you name it. AI browser tabs are open on virtually every laptop, whether in the classroom or on campus more widely. Surveys and reports have shown an increase in student usage, but I suspect the rates are even higher. I simply do not believe there are a significant number of students who have not incorporated AI into their working routine in some capacity. There may still be a solid number who do not use AI to “cheat,” however they define cheating, but overall usage is utterly normalized.

Second, students are absolutely aware of what they are doing, and rationalizations abound for almost everything. Cheating may be commonplace, but the definition of what is and is not legitimate AI assistance is messy, incomplete, negotiated, and will continue to be contested going forward. No one has any idea where to draw the line, and those who claim it's easy have probably not used the tools very much themselves.1

Third, faculty have adjusted to the new environment in wildly different ways, depending on school, department, and individual faculty member, and there is no uniform overall response except perhaps the recognition that the only work that can be guaranteed authentic is done in class without technology and in front of the professor. Everything else is up for grabs.

Fourth, unless something truly dramatic happens - some kind of major "token tax" that drastically reduces free access to AI platforms, for example - everyone in education is going to have to adjust to a new normal where students can cognitively offload almost any kind of knowledge-related work for the foreseeable future. Arguing whether AI is inevitable is now mostly pointless. It's akin to arguing whether the internet was inevitable.

Online commentators, instructors, professors, administrators, parents, pundits, and consultants need to accept these as the baseline. Many adults are still squabbling over whether students should use AI at all while the students have completely moved on.

The Shame is Gone

That students have been using AI heavily since last year is not really news. What my interviews revealed, however, was the extent to which college students have folded these tools into their daily lives. Of the students I spoke with, there was a wide range of attitudes on the proper use of AI in their academic work, but not a single person disagreed with my premise that almost all students are using AI to assist them. One student shared:

There used to be this shame around using AI for a lot of things. I noticed, at least in high school, people would feel like they had to be discreet when using Claude or ChatGPT or whatever. But now it's commonplace. It's commonplace in academia and it's also commonplace just among students. People are not shy about it.

Another, the student who used AI the least, described sitting in on a large lecture:

I mean, I remember one time I went to a politics class last year and I sat in the back and I could see everybody's computers or whatever. The range of tools that they use is so great, but almost everybody is using it. I remember one kid was literally transliterating everything that the professor said and putting it into a chatbot in order to get more detail … And I've walked through the library and just seen rows of people sitting there with ChatGPT open.

Even those students who are booing AI at college graduations2 and have strong opinions regarding AI’s environmental impact are nonetheless almost forced into using AI to keep up.

This from the same student above:

But yeah, I’d say AI use is absolutely normalized. People use it for every sort of discipline, for little tasks, for huge assignments, even in class. The worst one I saw was some guy asked ChatGPT what he should ask as a question in class, and then he raised his hand and said whatever it said. I mean, really, for anything and everything. So I think yeah, it’s definitely normalized. I would say on my college campus - I think college campuses across the country would find the same, very similar situation.

My thesis has always been that when surveys ask students whether they use AI, they typically hear the question as being about "unauthorized" use rather than any use at all. The student who uses AI to help them study, but would never generate an AI essay, will answer no in good faith. Most of these surveys were also administered a year or more ago, when using AI at all still carried more stigma. It's unlikely any of those numbers capture the reality of fall 2026.

I asked every student I interviewed whether AI was completely normalized on their campus and got an unqualified "yes" every time.3

What Counts as Cheating?

That said, the gradations, qualifications, caveats, parsing, and nuance about AI use were real. The students I interviewed acknowledged using AI in some circumstances and not others, for some assignments and not others, and in some classes and not others. Most were ambivalent about the technology itself. None denied using it. Some were heavy users and unabashed about using it to circumvent actual work. Others were careful to detail what was fair game and what was off limits. All of them pointed to what they considered reasonable use cases.

Some of those uses seemed beneficial. One student, among the hardest-working I've ever taught, made the case for AI as a tutor. I pushed him first on the objection I hear often from teachers, which is that the output can't be trusted.

Here’s his response (and just for context, this is a student who is not one to suffer fools - AI or not. He’s the type who will double-check anything he finds suspicious.)

So now I think most of these models have gotten to the point where if I'm taking an undergraduate introductory [science] class, I don't really need to provide it with source material. It knows its stuff without hallucinating. And the only reason I trust that this is true is because I've just seen this. Maybe if we're talking about really advanced topics and current research in the field, that's different. But we're talking about concepts that have been established for decades now, and I have a lot of trust in it.

He isn't worried about any of this hurting his learning - quite the opposite.

I don’t think it can ever hurt your understanding. And the reason I say that is because it’s really hard to cheat in these classes. At the end of the day, people realize that they’re going to have to learn the material. And AI is better at helping people learn the material, because you can ask all sorts of goofy questions, and you can work through answers, and it’s patient, and it’ll explain things however you want, fits your learning style.

I think AI explains concepts to me in a way that sometimes professors can’t, and I can’t always email a professor anytime I have some small question. So it’s actually perfect for that.

I use it a ton. Everyone uses it a ton, and I think they should. (emphasis added).

Just to be clear, this is a student who is as serious about his education as anyone I’ve ever taught. I was honestly a little taken aback at how much faith he had in the models given his skepticism from just a few years ago.

Many students were adamant that AI cannot think or write as well as they can, so that was where they drew the line. But even these students acknowledged that proofreading a paper was fair game. So was reducing a long reading to bullet points when time was short. So, according to one, was using AI to complete an entire online problem set. This aligns with Grant's student essays.

Another student described the evolution of his own use:

And I didn’t [freshman year] use AI at all, in the sense where I didn’t even use chat to revise my sentences. And my sophomore year, I found that I was an anomaly amongst my classmates, that most people were using it for cosmetic or editorial purposes at some level. Yeah, for editorial or even idea generation, although less so, just because again, at that time the creative functions for AI were not that strong. It’s gotten a little bit better with the deep research functions. Like, I use that a lot. I use it a lot for my research memos when I have to write them - it’s great at just combing data. But if I’m writing a philosophy or literature paper, I still don’t rely heavily on its creative functions.

One student whom I knew was also hesitant to use AI nevertheless pointed to generating practice questions and review sheets as a legitimate use.

Few used agentic tools or sophisticated workflows, and almost none needed paid models to do what they were doing. Many knew classmates who had paid accounts, but the consensus was that almost everything could be done on the free tiers.

Zero Faculty Consistency - But One Universal Trend

That faculty views on AI are all over the place is not especially revelatory. Most schools have delegated AI policies not just to departments but to individual faculty, and students reported an extraordinary range of responses. The humanities were generally more hostile to AI use than STEM courses, but exceptions were everywhere. One thing was consistent, however. Almost none of these students encountered any mention of AI during freshman orientation. Only one could recall anything at all:

There's a chance it came up in orientation, because we had such a long orientation that I wouldn't be surprised if they mentioned it at some point, but it was never in a way that stuck with me or I distinctly remember. But definitely in like the syllabi of most classes at this point, there's like an AI policy to some extent, or at least a mention of AI. So probably within like my first few classes, a professor probably mentioned it.

Policies are all over the place, mostly by design. Almost every school - K-12 included - has delegated them to the teacher. There may be boilerplate language from the institution or the department, but the reality these students described is that their professors are left to their own devices.4

Despite the individual inconsistency on providing guidance for students about AI, a clear trend emerged last year, put most plainly by a student describing an economics class:

My econ professor second semester, he said explicitly, since most of our problem sets can now be solved with AI, we're no longer weighting them as much as we do. They're going to be graded for completion and they're going to be like 10% of your grade, or 5% of your grade. And most of your grade will be exams, because you can't cheat on that. And if you're shameless enough to cheat on that, AI or not, that's not something that AI can make worse per se, because all of our exams are in person and proctored by like 20 people.

Not all professors are demonstrating the same level of awareness. One student described a 10-page-take-home report, requiring research, analysis, and an essay, assigned by a professor who said nothing about AI. The student was blunt - “Nowadays, because of how normalized it is, you have to address that.” He called this a “rookie” mistake.

These interviews captured a wide range of faculty responses to AI, but many supported the overall shift to higher-stakes in-class assessment.

Blue books came roaring back. Many teachers intend to double down on this strategy again this year but I wonder - is this a long-term solution or short-term survival? Are we really prepared to sacrifice the ability to assign any significant work outside the classroom?

What is the Next Step?

Given the reality of student AI use, it’s incumbent on educators to look for ways to inspire and motivate students to develop their own organic capacities, perhaps with a nudge from us, but ultimately they will need to arrive at that conclusion themselves. Several students pointed toward a path forward.

Here is one, unprompted, in the middle of a conversation about how his classmates handle online exams:

At some level, education is a two-way street. The student has to buy into their own education. It’s an investment of some sort.

He was not fixated on integrity. He was talking about value - what was college actually for?

Again, from the student most skeptical of AI, who by that point in our conversation had already told me he no longer judges anyone for using it:

It’s [college] such a privilege. It’s such a great opportunity as a student, regardless of what position or field you want to go into. It’s just such a shame if you approach that experience with the mindset of, “Oh, I’m going to try and make this as easy as possible so I have to do as little work as possible, end up with a good grade,” and none of it matters. I really think that’s a sad way to look at it.

Unless we can get the majority of students to hold this mindset, AI use will continue to pose a threat to the academic experience. Idealistic? Maybe. But why? Have we strayed so far that we now have to convince students they are in school to learn?

Apparently, yes. The way the current system is set up means this won’t be easy. But plenty of teachers have already built classrooms where it’s happening, and it’s worth looking into those rooms to see what they’re doing. I’ve been around long enough to see the difference between teachers who inspire, lead, and role-model and those who mail it in and practically invite cognitive offloading in their classes.

The students I spoke with made clear distinctions among classes, professors, learning environments, and assignments, and made up their own rules when they were and weren’t comfortable leaning on AI. They are figuring it out through trial and error. Before we render any sweeping judgment about their behavior, it’s worth finding out how and why they’re making those decisions. That’s the reality of college and AI in 2026, and it’s already drifting down into high school. We're five academic years into free access to powerful AI and most institutions are still in the same gear. Every year from here on out has got to be better than the one before it. We are running out of time.

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

Back in June, I ordered my transcript for a Masters program I completed online from 2021 - 2025. The St. John’s Great Books program is known for its rigor and unique reading list - not so much for its credentialing. The resulting de-emphasis on grading means SJC faculty only offer comments on papers and don’t share their marks with students at the end of each semester. The only way I could see how I did was to request my transcript.

When it arrived last week, I confess I was curious. Fifty-plus years in education - first as a student and now as a teacher - have conditioned me to accept the absurdity that a single letter grade represents a student’s entire learning experience in a class. In one seminar, we had no papers or other means of assessment - that grade might as well have been pulled from a hat.

Overall, I did well. Yet, none of those letters revealed how well I understood Plato, Kant, or the Bible. The Great Books program is seen more as a calling, a personal quest to actually read the classic texts with like-minded peers. Assigning grades in that context feels almost like an oxymoron, but as a degree-granting institution, it’s a requirement.

Once students reach a certain age, we issue letter grades for everything - papers, tests, projects, quizzes, presentations, and so on. The grading system is so deeply ingrained in schools we almost never question it.

AI has only supercharged the grading arms race. With competition for higher grades at an all-time high, the temptation to offload work, bypass learning, and produce solely for the grade seems higher than ever. The motivation to perform in school is almost entirely extrinsic.

The question we probably don’t want to know the answer to (or, the one deep down we already do) is:

How many students view learning as the primary goal of their time in class?

What Flow Requires

That question is really a question about motivation. The best book I’ve read about motivation was published almost four decades ago and isn’t actually about motivation.

Mihaly Csikszentmihalyi’s Flow: The Psychology of Optimal Experience is more relevant than ever to our AI-saturated world.

Flow is his name for the state in which a task absorbs a person so completely that it shuts out almost everything else - it results in pure enjoyment, even if the activity is difficult (like swimming the English Channel). While he identifies eight components of flow, three of them describe conditions a teacher can actually create. The activity has to demand real skill and a task with a realistic chance of completion. It must have a clear goal. And, crucially, feedback has to be immediate.

Every teacher knows instinctively that tasks - whether in the classroom, on the field, or on your own - need to be calibrated so that the difficulty level matches a student’s edge of capability. Design a task that requires skills students have not yet mastered, and the ensuing frustration prevents “flow.” If it’s too easy, boredom sets in. The golden ratio is that match between ability and difficulty that leaves us engrossed.

Establishing flow conditions in a classroom does not mean cheap entertainment or diluting rigor. In fact, the opposite is usually the case. Matt Brady, a chemistry teacher, opens the year by having his students make gunpowder. They have to convert percentages to grams before they can collect a single ingredient, and write it all up afterward. Teachers like Brady have always existed. As a student, I considered them eccentric. Now I know they were building the optimal conditions for flow all along.

In a section entitled “The Challenge of Lifelong Learning,” Csikszentmihalyi writes:

Extrinsically motivated education is a perfect phrase that captures modern schooling. The motivator is usually grades. Most classes students complete over the course of their academic careers are graduation requirements.

Csikszentmihalyi continues:

But a person who forgoes the use of his symbolic skills is never really free. His thinking will be directed by the opinions of his neighbors, by the editorials in the papers, and by the appeals of television. He will be at the mercy of “experts.” Ideally, the end of extrinsically applied education should be the start of an education that is motivated intrinsically. At that point the goal of studying is no longer to make the grade, earn a diploma, and find a good job. Rather, it is to understand what is happening around one, to develop a personally meaningful sense of what one’s experience is all about. From that will come the profound joy of the thinker …

A student who glides through school, even a successful one, but who has never been taught to question their own opinions or been challenged by someone with more knowledge, is utterly at the mercy of our current media environment. It accounts for so many of the problems with online behavior - toxic forums filled with misinformation, disinformation, hate, and ignorance as well as the superficiality of short-form video, influencer and TikTok culture. Teachers need to, unironically, model the “profound joy of the thinker.”

How Would you Get Someone to Actually Read This?

Human beings are naturally curious creatures - we want to learn new skills and new things all the time. Throughout the bulk of human history, the motivation was not external validation in the form of grades. It was something deeper, more primal, and fundamental, often life-saving.

My teaching log for my graduate program in the late 90’s included this provocative question:

Think back to a time when you had a high level of success in learning something that you feel was challenging to you. What was the nature of the learning situation that enabled you to master these difficult ideas or skills?

I wrote about my response to that question in a previous post, but I want to expand on the distinction between extrinsic and intrinsic motivation.

Motivation is essential when it comes to learning. The course I wrote about was one in which I became internally motivated. The instructor had set up conditions in which the learning - in that case, improving my ability to write well - overrode all other concerns.

For one class, the professor brought in a tool that looked like something from a medieval peasant village - it resembled a huge pair of ancient pliers. We were told to describe what it did1 in preparation for an assignment where we had to explain, in detail, how a common, everyday modern invention worked - something like a microwave or an alarm clock. I chose a stapler. My piece read like an instruction manual. I'll never forget his comment: "Okay, now how would you get someone to actually read this?" A lightbulb went off about the importance of an authentic audience, something never emphasized in my high school classes.

Often, internal motivation does not arise anywhere near a classroom. Remember a time when you tried to pick up a new skill for no reason except your own interest. Maybe it was a new language, an instrument, or a hobby like amateur photography. In those cases, the feedback was self-evident - could you talk to a stranger, play the chord, or take a picture that looked good? If so, you were learning. There was no grade. You pushed yourself solely because of a desire to learn and get better. I wanted to learn how to write a description of how a stapler worked that a reader would find worthwhile.

What scares me when I think about that course today is whether I would have handed the whole problem - observe the stapler, then explain how it works - to an LLM. I like to think that, because of the seriousness with which I approached the class, I would have abstained.

I have been told over and over by kids today that there is an easy solution to the cheating crisis if demonstrating whether the student actually learned the material is the true goal of the professor - all assessments need to be done in analog under teacher supervision. But that still mostly presumes external motivation.

On one level, that saddens me because I’ve also been told by students that when a teacher does rise to the occasion, when they make the course material come alive and infuse their classes with purpose, students want to do the work. They want to learn because it’s been made real for them. That was my experience in that freshman writing course. It is still possible.

Where I See It Work

Teachers who can genuinely inspire, coach, advise, and redirect students to master their area of expertise will become even more valuable in the AI era.

Despite so many stories about student cheating with AI and the lengths to which educators are going to prevent it, I wonder if we can look harder at our own practice. Thousands of teachers around the country routinely lead students to invest their best effort in their courses - whether it's algebra, chemistry, English, or computer science. As I frequently tell my debaters, if you are not convinced by your own argument, you can’t possibly expect to persuade someone else. If a teacher is not sincerely interested and invested in their subject matter, why should their students be?

In my experience, the teachers who tend to draw students in are passionate and highly trained in their field, model lifelong learning, are fair, and demonstrate care for their students. They are also, often, demanding. Even those qualities may not be enough to seed intrinsic motivation - what effective teachers do best is design the work differently.

In my own work, historical simulations and role-plays have had the most success turning the classroom into one where more students adopt an intrinsically motivated mindset. Students inhabit a real historical figure and compete in a goal-oriented activity with clear stakes and immediate feedback - the very conditions Csikszentmihalyi identified as conducive to flow. Those activities, often comprising a week or more, are the ones students bring up when they come back to visit years, sometimes decades later. It shocks me that I can often remember who they played and what they said. Simulations are flow classes for me as well.

I'm extremely fortunate. My class sizes are small, my school is well-resourced, and many of my students are highly motivated to begin with. Yet I still see the difference with these activities. I don't think most kids use AI to write their introductory speeches, position papers, or final reflections. I can't verify that with certainty, but the work is public - they deliver it in front of people who respond in real time - and the kids are invested.

What School is For

Csikszentmihalyi makes a provocative claim about schooling later in the book:

Similarly the value of a school does not depend on its prestige, or its ability to train students to face up to the necessities of life, but rather on the degree of the enjoyment of lifelong learning it can transmit.

The “similarly” refers to his observation that the value of a community should be judged on the opportunities it affords its members to live to their full potential. Thus, Csikszentmihalyi views school as a place to cultivate the importance of learning above all else. Most educators, and Csikszentmihalyi himself, would agree that’s not happening in our current system.

Others might argue that school is a place to obtain skills first and foremost for their entry into society. Why can’t they do both? Why can’t students experience the joy of learning and acquire the necessary skills for work and for life? Too many students experience school as a place that extinguishes the joy of learning rather than celebrates it.

Most of us got into this because learning is not only beneficial and meaningful, but can be profoundly rewarding under the right circumstances. Summer is the time to remain idealistic - I know what happens when the reality of October hits. But as we walk back into classrooms over the next month or so, it's critical to remember that our job as educators is to create, preserve, and sustain those conditions for learning. Hopefully, even with the ability to produce answers and bypass flow just a click away, more students will experience the joy of learning intrinsically in classrooms explicitly designed for it.

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

I’ve started and restarted this post a dozen times as I’ve gotten ready for hip replacement surgery this summer (which finally happened last Wednesday!).1 Experiencing the detailed care of my surgeon, nurses, and physician assistants only reinforced how much their protocols run on technology. Visits are often recorded, images documented, and almost every stage of the process digitized. When I asked whether any of this data was integrated with AI, the short answer was yes - with the privacy and security caveats you would expect. The hospital was the latest reminder that AI is being folded into every industry we encounter, whether we know it or not.

Everyone, including teachers and students, needs to work in an AI world. Almost every story about AI and education invokes the risks of cognitive debt, cognitive offloading, and a loss of cognitive patience. As we enter the 5th academic year where students have had access to powerful AI, most of the news out of schools is negative. More telling, even the most charitable observers would agree that few of us know what we’re doing when it comes to AI. But I’m hopeful that will change this fall.

I've read the reports and documented my AI fears when it comes to unsupervised student use. What I'm trying to square is that my own experience with AI has been mostly positive. Working with AI over the past three and a half years has required a learning curve, but it has made me a better planner, a more precise writer, and it has forced me to slow down and review my work much more intentionally. With billions of people now using AI models, it's clear I'm not the only one benefiting.

So much of the angst around students' AI usage presumes that they will continue to use it the same way they have over the past three years. We have to remember it is still early days. Getting real benefit from AI takes a classic set of skills (or habits or dispositions if you prefer), and every one of them is something we have always taught, long before 2022. Effective AI use requires planning, precision, and patience. They apply whether you let a chatbot into your classroom or not.

Learning Skills Are AI Skills

One major problem for anyone advocating student use of AI - whether in the classroom or on their own - is that there is very little evidence it helps students learn. The further I get from November 2022, the more it shocks me that this technology was unleashed on an unsuspecting public, especially teachers and students, with so little warning. No other “arrival technology” in my 30 year career has disrupted the teacher-student contract as much as generative AI.

I’m aware that my usage rests on skills I formed over decades. Yet, I’m confident we will ultimately find ways to absorb AI into education by leaning on the same meta-cognitive habits we have always valued. Planning ahead, communicating with precision, and having the patience to see a project to completion are all things any teacher in any classroom already models.

The critical question over the next several years is when and where students will continue to build these skills - alongside AI assistance or away from it. Right now, most teachers lean towards the latter, and that may hold for a while. What I don't see, however, is how we are going to have a choice regarding student usage outside of school. My hope for this fall is more evidence of educators successfully modeling these skills, regardless of whether they bring AI directly into the room with their students.2

What Do People Do All Day?

I often share with students the reality of work. Years ago I read that college-bound high school juniors are as busy as CEOs. There is almost no other time in your life when you will be expected to excel at five entirely different subjects. Most of us will eventually get the opportunity to specialize in an area where we are strongest, but that is usually not the case in high school. That fact alone explains a great deal about how students use AI. A kid managing five courses plus outside school commitments has no time to plan, no incentive to be precise, and little patience left for a second draft.

I've explained to students that almost all work is best approached by breaking it into smaller parts. To do that, you need to check your anxiety, step back, and survey what actually needs doing.

AI raises the stakes on that crucial habit. We have all gone down a rabbit hole, and AI rabbit holes are among the worst, partly because the machine flatters you at every turn and will happily lead you somewhere useless or overly complicated if you're not careful. The discipline any project requires has to be in place before you ever go near a chatbot. If you know what you're doing, though, AI can assist with the planning by challenging assumptions and expanding on strategies. It can also help you decide if the task is worth using AI on at all.

Planning

My hunch is that a startling number of people still use their chatbot as a string of disconnected one-offs. If you use AI infrequently, you probably turn to it for a few questions, close the tab, and move on. Which means you have dozens of orphaned chats piling up in a sidebar, unnamed and unrecoverable. If that is still how you work with AI, you are missing what's been happening over the past two years.

I have no interest in writing an AI tutorial - there are dozens available online. But the models have rapidly added organizational tools and none are difficult to learn. Keeping related chats inside a dedicated Project folder means the work stays together and the model can draw on related material. Adding a set of instructions and reference documents within each folder considerably improves the output. These are simply AI versions of separate notebooks and divider tabs, a habit older than most teachers. Gemini, ChatGPT, and Claude offer almost all of these for free, and the lowest-tier paid versions provide access to even more sophisticated capabilities.

We constantly assign students work to do, but in the age of AI, we have to remember that planning out a multi-step activity - whether it is a research paper, a science project, or a presentation - is a meta-cognitive process. Is the goal to produce a document or ask for feedback? Are you creating, analyzing, or organizing material? Is this for you, or for someone else? Nine times out of ten, simply asking yourself what it is you are trying to do will force a decision about whether or not to use AI in the first place.

None of this is new to teachers. We have explained and modeled executive functioning skills for years, through graphic organizers, outlines, and day planners, well before AI was available. I am a longtime evangelist for David Allen's Getting Things Done, and I wish someone had handed it to me before college. Planning ahead does not come naturally to the teenage brain. It never has. That is precisely why we teach it. We need to prioritize the planning before the AI, now more than ever. Refusing to recognize that AI can help with the planning itself, however, is also a mistake. As adults, we must constantly model judgment.

Precision

Using AI has made me a better writer for one simple reason. It’s made the cost of vagueness instantly painful.

I've written previously about growing up in a household filled with lawyers who were maddeningly fixated on using precise language. It created an ideal training ground for writing clearly. Law school provided three more years of practice. Writing with precision and clarity was not an abstract concept in my household - I saw it up close in every brief my father ever showed me.

Iterating a conversation with a chatbot is an exercise in specificity, context, and requirements. If you want to convert notes from a meeting into a bullet-pointed to-do list with each entry no more than five to seven words, then you need to describe those specifications. Or if you want to design a research prompt that will focus on retrieving publicly available academic, government, and other vetted documents only, you need to ask.

When you do not precisely identify what you want, when you do not supply the context of the request, when you are not even sure yourself what you're after, you should not be surprised when the output is subpar. The more detailed and clear your request, the better the results. Whenever I write a prompt, I am constantly revising and scrutinizing my work to ensure there can be no misunderstandings. Most of the disappointing output traces back to an unstated assumption in my request, or some instruction I didn't even realize mattered until I saw how the AI responded.

We do the same thing with students. Using AI has reminded me how much depends on clear instructions. We cannot blame the student for misunderstanding an assignment when we don't take the time to be specific.

This is also true from the student side, and one reason that modeling precision is the province of every writing teacher. Anyone who grades student work has written some version of I don't know what you mean here in the margin - writers often forget that the reader is not inside their heads, and younger writers are especially prone to it. Using AI delivers feedback on fuzzy requests immediately and without mercy. The truth is that a teacher who never allows AI in class is still teaching the skill that matters most if a student opens a chatbot on their own.

Patience

Most kids just want to be done with school, period. They want the current assignment off their plate and the very thought they need to “do it over again” induces horror. Anyone who works around teenagers knows this.

If you put the same kid on the field, however, something surprising happens. They will run the same drill over and over until their coach stops riding them. It’s not that they don’t understand what patience looks like. It just doesn’t map onto their experience with reading, writing, or arithmetic.

I am a firm believer that writing is rewriting, and what is rewriting if not patience? When a student wants to hand in a first draft, the question I ask is whether it is their “best work.” Most of the time they hesitate which tells us all we need to know. Our first drafts are never our best work - adults included.

None of this has anything to do with AI. Patience is what we model every time we tell a class that something polished - a poem, a painting, a song - took longer than you might imagine. Extraordinary results require extraordinary patience.

This is why the way AI often gets demonstrated to the public bothers me.

Ethan Mollick recently dropped another one of his “state of the union” posts on the current abilities of the top models. His work is the most up-to-date and accessible I have found, and I recommend it to anyone who wants to keep track of AI developments.

One area where I think he does his readers a disservice, though, is his demonstrations of impressive and "cool" applications built from a single prompt. In his most recent post he compared one line about a procedural brutalist building creator across two models, with the newer version significantly "better," if better means it looks cooler than the first one. There are a lot of things AI can do that are "cool" - but is that what makes it useful?

I understand his larger point, and the mere fact that these models can do such things, considering where we were even a few months ago, is mind-boggling. But no one doing serious work with AI does so based on a single sentence prompt.

The way AI is going to change work is through the people who figure out how to tease a project apart, separating the human review from the repetitive labor AI excels at. When those people begin to pull ahead of the people who either do not use AI or use it ineffectively, then we will see its impact. That takes the patience to iterate a project from the beginning and see it through, rather than issuing a one-shot ten word prompt.3 Perhaps we will get to a stage where that becomes the norm, but we are not there yet.

The Year Ahead

Two years ago I was fixated on simply being AI-Aware as we were all learning how to handle the emerging AI crisis in schools. Last year my strategy was AI-Proofing, which was defensive by design, as I, along with many others, attempted to recreate assignments that could not be easily outsourced. The evolution makes sense, but it's not where I think we will end up.

My mindset this year turns to instruction - not specific to AI, but a re-commitment to the durable habits of mind that improve learning no matter the tool. By reminding students of the importance of planning ahead, communicating clearly, and exercising patience in everything they do, we will have a successful year -with or without AI.

I have seen all three of these skills converge this summer as I have been building out full websites for my courses.

The planning came first, as I had to think carefully about what I actually wanted these sites to do, what information they needed to contain, and how they were set up. Precision was present throughout, as I needed to specify exactly what I wanted each page to do, where and how I wanted the links to operate, and what kind of layout and formatting would make it most useful for my students.

Patience has been ongoing, as I am still iterating and fine-tuning before the school year starts. A one shot prompt might have created something "cool" but would have produced garbage as far as my actual classes were concerned.

The web design project required skills applicable to any task - this one just happened to incorporate AI.

The Wrong Argument

Hospitals, marketing firms, technology startups, and a whole host of other industries have leaped feet first into AI. No one wants to be left behind. Results have been unsurprisingly mixed even though massive amounts of capital have been invested. How this all shakes out is anyone’s guess but we won’t know for awhile. Amara’s law hovers in the background - “we tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.”4

Education is still behind. Few teachers are taking advantage of the most powerful ways AI can be used to augment their work. Most of them are likely keeping it to themselves or, occasionally writing about it publicly. Many teachers (unlike their students) are still deciding whether AI should be allowed on campuses at all.

That argument, I think, is the wrong one to be having. Teachers who want to use AI with students and teachers who want nothing to do with it can model the same habits, and both can point out that those habits are what separate using these tools well from using them badly. In both cases, student walk away with something that works. Technology changes. Good teaching does not.

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

One of the perks of a teacher’s summer is the pile of books we now have time to tackle. I estimate that I’ve spent upwards of six hours a day reading over the past twenty-five years. That’s not unusual - I suspect many of my readers engaged in knowledge work exceed that count. A rough calculation puts that at more than 54,000 hours of reading. Using a modestly high reading speed, that easily translates to more than a billion words over that time.

This means I can tell at a glance whether a Google AI summary is likely to be on target, whether I need to read deeper on the topic, quickly locate a trusted article or source to skim, zero in on the section I need, or whether I should regroup entirely and start a new search. I do this effortlessly, without much thought, and feel confident about the process. Most students can’t. Students will not or cannot read AI output with the understanding or the skepticism it demands, because most do not have anywhere close to a billion words under their belt to parse its language. The overlooked issue is that if we actually want students to use AI effectively, either in class or at home, the skill we need to worry most about is not writing. It is reading.

How to Read a Book

Last summer I argued that the teaching method in the Great Books program I finished was the one AI can’t touch. One flaw in that argument is if students start off-loading their reading to AI. But even if they do, many students do not have the ability to read AI summaries carefully, let alone ones based on classics like The Republic or The Odyssey.

A book I had heard of but never read before the St. John’s program is Mortimer Adler and Charles Van Doren’s classic How to Read a Book (Van Doren of Quiz Show fame). In it, the authors lay out a method for tackling more advanced texts. The whole book may seem dated, but it is worth reading, especially the opening chapters. My major takeaway is that we rarely teach more advanced reading skills to improve comprehension once we have finished teaching basic decoding in the early grades. More teachers would benefit from applying some version of the strategies they recommend.

A piece by Carl Hendrick sharpened my own thinking on the critical importance of reading.1 A prompt typed into ChatGPT, he points out, contains almost none of the knowledge in the answer it produces. The model’s response comes from everything it absorbed during training. Human readers operate in much the same way, supplying meaning from a lifetime of accumulated vocabulary, background knowledge, and life experience. The difference, of course, is that while AI may have access to billions and billions of words on virtually any topic imaginable, what it doesn’t have is emotions, experiences, or human connections to the world that provide uniquely human interpretation when reading a text. An AI cannot be moved to tears at the conclusion of a novel even if it is able to identify the emotionally charged language embedded in a paragraph.

When students do not have the foundation built through thousands of hours reading on their own, they do not have the vocabulary or muscle memory necessary to impart deeper meaning, especially as texts get more complex. Unsurprisingly, reading is one of those activities which improves exponentially - the more you read, the better you read, the more you understand, and the better you read and understand, the more you want to read.

Having read more than a billion words is a by-product of my disposition and my profession - it provided the training corpus that lets me read as efficiently as I do, and students simply haven’t put in the time. That is one reason why AI’s fluent and confident output often goes unchecked.

As a history teacher, I am often guilty of the very thing Adler warns about, even though I have tried to make reading instruction more a part of my practice and feel even more compelled to do so in the age of AI. I assume kids read better than they do, and I am programmed to spend my attention on content. That assumption needs re-evaluating, especially as we head into a world where delivering content instantly, on demand, synthesized, and drained of its original meaning becomes commonplace. We still need students to handle literary fiction, dense secondary sources, and primary documents.

Different Kinds of Reading

Not all reading is the same, which is a crucial thesis of How to Read a Book. Reading a thriller for its plot is a different act from reading a novel for its ambiguities, which is different from reading a work of history or philosophy for its argument and its evidence. Many people read mostly for pleasure, passively, like watching a film. Few people want to struggle with more difficult texts on their own. Most students, even many of our most proficient readers, rarely tackle tougher texts independently. Online social media rewards short paragraphs, skimmable listicles, and now, because of changes to Google’s search, instant answers that arrive before the question is even fully formed.2 Maryanne Wolf calls the capacity we are losing cognitive patience, the willingness to stay with a difficult passage long enough for its meaning to surface.3 Deep reading has always been the hardest level to sustain and the easiest to lose - even adults frequently report difficulty staying engaged in deep reading in today’s social media saturated environment that rewards and cements ever shorter attention spans. Any child in school today is going to exist in a world where instant answers will be the expectation, not the exception.

Evidence from the Classroom

One of the best courses I ever took, for my Master’s at Columbia’s Teachers College in the late 1990s, was titled Building Bridges and Breaking Down Walls. It was co-taught by two public high school teachers, one in English and one in History, in Manhattan. We met for three hours on Saturdays over twelve weeks. Unlike the education courses I took over the summer, where we read Foucault or Derrida or some other theorist at a fifty-thousand-foot level, I was teaching full time while enrolled in this class, so the fact that it spoke directly to the realities of the classroom allowed me to implement what I learned the following week.

One point that resonated was how English teachers tended to emphasize the language and intent of the author, while history teachers were more concerned that kids understood the content. But the overall focus of the class was the critical importance of literacy in general.

Read it Again

The English teacher related how, early in his career, his reflexive response when a student expressed difficulty with a text was to tell them to read it again. This happened several times, in the same class, with the same student. Each time, he simply said read it again. Finally it dawned on him that the “read it” part of the instruction was where things had gone sideways. The student could not read it in a manner that allowed for comprehension. He had been treating a reading problem as a motivation problem. It finally occurred to him that he needed to explicitly teach reading strategies.

That story has stayed with me because it matches my experience - most high school teachers assume, though rarely examine, that students can already read well. Most of us were trained in our subject, not in the pedagogy of reading. Much has been written about students’ poor reading abilities, some of it designed to go viral,4 but almost always in a helpless, throw-up-our-hands register or scathing dismissal of “kids today”. We are much better at observing the problem than proposing solutions.

I began folding the reading strategies we discussed into my middle school history classes. I built a handout of increasingly difficult passages, and we annotated with marginalia - text to text, text to world, and text to self reflections - along with questions, words or phrases that puzzled us, and whatever we could untangle behind each short section. The point of the lesson was that there are many levels of reading, and you should not feel ashamed or inadequate when you hit an impasse. Friction and struggle were the point.

I believed that because it described my own reading life. I have read everything in front of me since I was ten or so, including many books well over my head. All of that time invested built whatever persistence I have as a reader. You do not need to understand everything to benefit from a book.5 Struggling with a text you cannot fully grasp stretches the muscle and introduces you to ideas, vocabulary, and situations you have to work through yourself.

With students, we need to help them with more advanced texts. Luke Morin has written about watching his students flounder with a single demanding sentence from Ray Bradbury, not for lack of effort, but because they did not know enough of the words to make a coherent interpretation.6 Past a certain point, struggle stops being productive. Students shut down. Assigning useful books is largely a matter of calibration - and sharing your own struggles is a wonderful way to model for students both the joy and the challenges of working your way through a difficult text. One of my most enjoyable teaching experiences in my entire career was reading Plato, Aristotle, and Homer with my Great Books students this past year. I tried to emulate my St. John’s tutors, who offered conditional interpretations and invited students to share their views. Reading a challenging book together can be a magical experience provided everyone is equally invested.7

Reading in the Age of AI

AI has turned reading on its head. The overwhelming focus in school, online, and in faculty meetings is on the impact of AI on writing. That is understandable, since writing is what we grade and what we require as proof of learning.

Reading, though, is the unexamined half of the literacy question, and it is where I think we should be more focused, especially as more schools bring AI into the classroom for students to interact with directly.

Marc Watkins identified this early, back in 2024, in a post bluntly titled “No One is Talking About AI’s Impact on Reading.” He warned that our fixation on writing was letting the reading question slip by unnoticed. Two years later, with AI far more embedded in how students find and process nearly everything, the focus on reading continues to be eclipsed by coverage of AI’s impact on student writing.

The irony buried in all the attention to writing is that reading and writing have never been separable. Ask any writer and you will find a voracious reader. A student who does not read will struggle to write well, no matter how many outlines and graphic organizers we hand them. Schools need to spend as much time thinking about teaching reading as we do about writing, and many of the ways AI is being marketed to help students read are just as controversial as its writing assistance.

Should We Individualize Texts for Ability?

A few years ago I used Brisk8 to help a student learn important material, and the decision perfectly captures the skill-versus-content dilemma identified in my Building Bridges course. I had a tenth grader who could not summon the energy to read in class. During silent reading and annotation her head slumped onto the desk, and she told me, repeatedly, just like the student in my teacher’s story, that she did not understand it. It did not help that the text was a hard secondary source on why the new United States had to abandon the Articles of Confederation. It was filled with abstract terminology - federalism, states’ rights, separation of powers. I needed her to understand the content for the next assignment.

Between classes, I used Brisk to rewrite the material several grade levels down, so she could work through it with our learning specialist. It helped. She understood the basics and was able to participate in the next activity.

Was that the right use of AI? She got the content she would otherwise have missed. The personally tailored version of the reading did its job and was not something I could have replicated easily. On the other hand, the simpler text removed the friction and the struggle. Did that make her a better reader?9

I solved the content problem by skipping the reading problem. But if we multiply that single decision across every assignment in every humanities class, what happens? When every student can dissolve any hard text on demand, where do they have an opportunity to build those crucial skills?

My very first Substack post described an AI exercise I did with students in which I built a custom GPT trained on a chapter from our American history textbook. The key instruction was that students had to read the chapter first, then use the GPT to go deeper and ask questions. Part of the activity was also to hear how they felt about using AI this way.

Looking back, I thought it was an innovative use case, but it’s telling that I have not done it again. I am not opposed to using AI in select circumstances to support reading comprehension and probe a text more deeply, but I was not convinced students would actually read the original first if they knew AI was waiting to help with the analysis. As I said in the post, I would never want students using AI in place of reading the original, but, just as with AI-assisted writing, the risk of reverting to cognitive offloading was too high. The other takeaway was equally clear: the strongest students got the best results. The ones who already read and wrote well quickly figured out how to improve their prompts and, critically, actually read the output with care. That’s something every educator needs to remember whenever AI is brought into the classroom.

As students get older they can find ways to extract AI’s benefits - and there are many - but I am more and more convinced that the bottleneck to effective student AI use is mostly a problem of reading. The sycophancy and confidence of AI output - worsened by the fact that students rarely know sophisticated prompting techniques or have access to the most capable frontier models - means many are not getting the most effective AI experience. And models carry worldviews that almost no one is aware of - a recent piece in the Economist investigated the many ways frontier AI companies shape the values that show up in open-ended responses, according to their own goals and preferences.10

If teachers do use AI in class, they need to think hard about how to structure the interaction to reduce these tendencies. Whether that means asking the AI to respond at an age-appropriate reading level (you can specify the lexile band) or, better, capping each response at 250 words or fewer, the average student user is otherwise likely to be overwhelmed. These workarounds exist, and I am sure many teachers have found them, but from my vantage point, and from what I see in the media coverage of AI in classrooms, almost none of this is discussed.

We Have to Be More Deliberate About Reading Instruction

I say this as someone who considers himself a strong reader, which is what makes my worst school experience instructive. I was terrible at science, in retrospect, largely because I could never get through the textbook. The vocabulary stopped me cold and I never built the discipline to look up each term. When I did, it only led to more technical language, and with no context I understood, the definitions themselves were impenetrable. This is exactly what Hendrick describes. I could “read” as well or better than my peers but had none of the domain knowledge, so I could not puzzle through the passages. It is the same wall plenty of confident readers hit the first time they try to read a dense Supreme Court opinion, and the one our students hit daily in classes where we assume they can read but the content is completely foreign.

Legal reading was the mirror image of my science debacle, though not because it came any easier. Motivated no doubt by early exposure in a family filled with lawyers, I found that my favorite part of practicing law was research, though learning to read a legal opinion still took real time.I often forget that this did not come naturally to me either. It was built from poring over thousands of documents over years - annotating, analyzing, questioning, and writing about the issues. Legal training in many ways can be summed up as learning how to read like a lawyer.

The Pitfall of the AI Summary

Relying on AI summaries without ever examining the source is a serious problem. The difficulty of research today is no longer finding sources - an AI Deep Research report can surface more relevant material in minutes than most of us could in a week before the advent of AI. The difficulty is reading that report, with its hundreds of citations (many from Wikipedia or Reddit unless you write a rigorous prompt restricting it to academic journals and other vetted sources), and doing the real work of tracking down the best ones and reading them carefully. How many students - how many adults - actually do that? Driving home the importance of careful reading is going to be among the most critical things we teach in the coming years.

Reading is so essential to almost everything we do in school that it’s sometimes easy to forget just how important it is to be a fluent reader. It’s the cross-cutting skill for every subject. In the Age of AI, it may be the most overlooked, and most valuable, habit we can model and emphasize for our students. If I could wave a wand and instantly bestow one proficiency on every student I teach, it would be the ability to read deeply. So many doors open once that muscle is built. Reading carefully and critically may be the single most important weapon we can give students when it comes to using AI.

To be clear, I am hardly a reading specialist. I have taken a few graduate classes, read some books on the subject, done some professional development, and, most importantly, worked with kids for over thirty years. I am more and more convinced that reading is what students most need, and that, for most secondary school teachers, it is something we are least prepared to teach.

There are many experts on Substack, but the two previously mentioned - Luke Morin and Terry Underwood - have influenced my thinking more than most, and both know far more about literacy and teaching reading than I do.

Terry Underwood has forgotten more about literacy instruction and research than I will ever know; go back through the archives of Learning to Read, Reading to Learn if reading pedagogy is a passion of yours. As I was finishing this piece he published a characteristically sharp case for how AI, used well, can help reading instruction.

Luke Morin’s Middle School Literacy Project is one of the most useful and humbling repositories of reading strategies I have found. And he writes beautifully.

If reading about reading is your jam, their newsletters are a goldmine for learning more about reading instruction.

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

Since school ended a few weeks ago, I have pulled back a bit from the AI conversation. I needed to get outside and away from my computer after a very busy spring. The World Cup has helped, though not as much with the getting-outside part. The rule changes have improved the viewer experience and the U.S. Men’s team is onto the knockout stage, which is a bonus.

Watching the matches is a reminder of the first time someone demoed ChatGPT for me. It was the first weekend in December 2022, the weekend the U.S. lost in the Round of 16 to the Netherlands in Qatar. I was running a debate tournament. We had a big screen up to display the pairings, and another coach pulled up ChatGPT, which had debuted a little over a week earlier. I hadn’t heard of it. I asked it to write a poem about the World Cup in the style of Shakespeare, which it did, in seconds. I was amazed. What an innocent time!

After millions of words written about AI and schools, here in the summer of 2026, I think students and teachers have arrived at something like a detente. If we intend to move forward productively in the fall, we need to stop fighting about AI long enough to ask what school is actually for.

The AI Arc (Late 2022 - Summer 2026)

Generative AI arrived out of nowhere, and no one was hit harder, day to day, than teachers and students. Amazement was immediately followed by fear - mostly of cheating. For some it generated excitement, then anger and backlash as people learned more about how these systems were trained, what they cost, and their environmental impact. A 2026 Brookings report on K-12 found that, on the current trajectory, the risks of generative AI in schools outweigh the benefits. Skepticism was present from day one - both about the technology itself and even more about predictions of AGI, mass unemployment, and the end of humanity.

Ironically, most of the apocalyptic messages come from the people closest to the technology, who continuously tell us it will eliminate jobs, surpass human intelligence, and remake society in ways they themselves describe as dystopian. The founders seem genuinely surprised by the backlash, even though the frustration so many educators and the public feel about AI is a rational response to being told the technology will upend their lives.

For students, access is now ambient. Anyone with an internet connection can pull up ChatGPT or their favorite model, which is orders of magnitude more powerful than the ones that shipped in 2023. Every assignment now carries a decision for them - how much AI to use, or whether to use it at all. Many are already dependent. But plenty remain wary, and more than a few are frustrated and angry that they were placed in this position without being asked. On that, teachers and students are often in agreement.

The AI question has both prompted and been folded into a larger conversation about phones, laptops, and the screens we have handed children going on two decades.

Schools, caught in the middle, have gone in wildly different directions. Some went all in, embedding AI into their infrastructure and their mission. Some refused and banned it outright, and many are still on their heels, yet to engage with the issue head on, churning out increasingly convoluted AI policies they have no way to enforce.

Most landed somewhere between deer-in-the-headlights and reluctant adoption, pushed along by district and university partnerships with AI companies often made under pressure to do something and not be perceived as falling behind. A few have taken thoughtful steps to confront the situation as it actually exists, resisting both the hype and the doom. And the gap between the people who use AI daily and the people who either reject it altogether or hold it at arm’s length keeps widening.

The school AI conversation sits in a larger AI media ecosystem that constantly veers from one pole to the other, making it even harder to focus on what matters amid the predictions, prophecies, and promises.

That’s the landscape I see as most schools close out the 2025-2026 school year. We’re battered, exhausted, but still standing.

Stalemate, or Detente?

My gut initially told me students and teachers have reached an AI stalemate this past academic year, like two prize-fighters circling each other in the ring, each exhausted by the flurry of punches thrown in the opening rounds. Now they’re stumbling, hoping the ref will step in and call a draw. Students cheated with near impunity for the first couple of years, and teachers fought back with invasive detection software and a return to blue book exams. Honor codes were rewritten, disciplinary outcomes were murky, and most accusations weren’t worth the time-sink to pursue.1

But this is what happens in an arms race: no advantage remains for long. Detection software was answered by humanizers built to defeat it, which were answered by better detectors, which were answered by better humanizers. Every advance invites a counter. Escalation with no decisive edge eventually grinds everyone down.

I don’t think student use of AI fell this year - the statistics tell the opposite story - but my hunch is they were more likely to use AI on the margins, where teachers could not or would not enforce strict “no AI” policies. Professors began, some enthusiastically and others reluctantly, to allow AI in limited circumstances, signaling not surrender so much as a recognition that AI had become an ever-present temptation to offload their work. Many instructors began to think more carefully about what they were asking students to do and why - a much-needed return to designing assignments around the authentic learning we claim to want. If AI leads to more curricular reevaluation that is, on balance, a net salutary effect.

What the Conflict Is Actually About

The detection war looks like a draw. It has created something closer to a detente - a condition where both sides are ready for a path forward. Students, repeatedly and consistently, have told us they want to be taught by human beings. Strip away the panic and that is what remains. The teaching relationship has always been at the heart of education.

No student ten years from now is going to tell a story about the chatbot that went off-script to tell them where it was on 9/11, or the chatbot that kept them after class to cheer them up, or the one that talked them off a ledge before a hard exam. Chatbots cannot read the mood in a room for those “teachable moments,” which are the ones students remember most. This is especially true in K-12 settings.

This is my essential observation. When that 2026 Brookings report concluded the risks of AI in schools currently outweigh the benefits, a central harm it named was the erosion of the relationship between students and teachers. The thing the research worried we were losing is the thing students keep telling me they want most. Whatever AI we let into schools has to protect that first.

This doesn’t mean AI has no role or cannot be used in ways that play to its strengths. Research is already being transformed by AI, and the ability to build bespoke tools through natural-language coding has real potential.2 But the notion that AI should become a central presence in U.S. classrooms, delivering and teaching content independently, has been undermined, especially over the past year.3

This is from my vantage point. If you are a teacher whose AI integration is really working, I’d love to hear about it. I’ve seen some of my colleagues use it creatively, and I know practitioners who are using it thoughtfully.4 But these examples feel like the exception rather than the rule, at least at the moment.

And despite what edtech vendors are selling, I have come to think most AI wrappers built for the classroom are a waste of time. For students who turn to AI to avoid their work, removing friction is the entire point; a wrapper that adds it back is a tool they won’t use. For the ones who want to use it well, they don’t want AI-lite any more than someone handed the keys to a car wants to be handed a bicycle. If we really want to teach older students how to use AI, they need access to the frontier models. For middle school and younger, classroom chatbot use is an entirely different conversation.

None of this means the detente is stable. The student who blew a deadline and is staring at a blank page at midnight on Sunday will still reach for AI, at least to get started - and no strong teacher-student relationship or coherent policy will reliably stop them. And the technology will not stay still long enough for us to get complacent. I have no idea whether the next disruption involves the proliferation of agents that move past chatbots, or ubiquitous smart glasses, or world models that create issues we haven’t even thought of. Anyone who tells you they know where all this is going is practicing a form of divination.

Student Confirmation

I’ve carried this feeling for a while this year, and recent conversations with students returning home from college confirm it. The former students I have talked with this spring keep saying versions of the same thing: most try not to use it but either can’t help themselves or have found specific tasks where it genuinely helps. More than a few have told me that they feel least guilty about using AI when they have a professor who doesn’t teach the material well in the first place.

Marcus Luther, a high school English teacher who writes really thoughtfully about teaching on his substack, The Broken Copier, asked his students again this year whether he should bring AI into his class. 81 percent said no, with only 6 percent in favor. One wrote that using AI would make them feel “way less accomplished.” His tenth graders, asked whether they wanted more or less technology in class, went around the circle saying the same word. Less. Less. Less.

While I have had success with AI tools in my advanced classes, I find myself landing closer and closer to his conviction. Using AI well in a classroom takes real expertise that most of us do not yet have, even as students keep using it everywhere else. We owe them guidance for the world they're heading into, but I'm not convinced how much time that means spending class on AI itself. That’s one of the issues I wrestle with the most. I don’t know how representative his experience is, but his student’s reactions don’t surprise me.

I imagine some of his students use AI, but my guess is not in there. He has built a culture of caring, noticing, and authenticity that makes kids want to learn how to write, how to think, and how to do the work themselves. I suspect I would have felt the same way in his class. The trouble is that not every student gets a teacher like Mr. Luther.

Angela Duckworth has described the same deep ambivalence in young people more broadly, many of them worried the technology is making them “lazy and stupid.”5 The pattern is consistent, and recent articles have named the trend: those using AI the most are often the most critical of it.6 This is the same impulse behind students booing commencement speakers this spring for invoking AI. Students are using AI more and understanding it better, but a growing number do not feel good about what it is doing to their learning.

We still do not have strong evidence that AI improves learning. A 2026 Oxford University Press study found that 72 percent of UK pupils, given the choice, declined to use AI on a school task, even as a majority used it weekly at home.7 One pupil put the distinction better than any researcher could: it “wouldn’t be what I truly think - a vague interpretation of what I’ve told a robot to expand on.” That absence does not prove AI has no place in school, but it does shift the burden onto the people promising the gains. I have found it immensely useful in my own work but have had mixed results using it with students. Talking about it with them helps. Experimenting is the only way we learn what’s effective.

More and more, I think our role in the classroom is to double down on the uniquely human connection, comfort, and guidance we provide. My ideal future is not one in which AI is absent from schools but one where students and teachers decide together what school is actually for, and only then where AI might have a place in it. I believe this will be one of the most important questions for educators over the next decade.

Why the Analogies Fail

I’ve never liked the printing-press and calculator comparisons. I wasn’t around for the printing press, and I barely remember the calculator “wars.” But AI is different in kind from other technologies that have reshaped schools. It strikes at the very heart of the learning process: it generates information instantly, answers the kinds of questions schools tend to ask more efficiently than students can, and easily undermines the very skills we want them to build. At the same time, and I can’t stress this enough, the technology is useful in the right hands and has the potential to let people do some increedible things. I reject any frame that refuses to hold both of these realities at once.

This is not an anti-AI piece. It is a pro-human one. We cannot keep bemoaning the problems AI brings without also examining the opportunities we have as educators to recommit to first principles. It means having the conversations we have spent four years avoiding - between teachers and administrators, between teachers and students, and among administrators, teachers, and parents. If we can do that, the past four years will not have been wasted. They will have set the table for the work that comes next.

A Final Indulgence

Before I published this piece, I asked ChatGPT to write a poem about the U.S. chances in this World Cup, in the style of Shakespeare - essentially the same prompt I used in 2022. Here is a snippet.

Enter Fortune, masked as VAR8

I am that glass-eyed witch whom none can love,

Who halts the shout, delays the leap, the kiss;

I pluck the eagle by the wing above

And whisper, “Was his shoulder off by this?”

The Americans are brave, and bravely young;

Their legs are thunder, yet their hearts must learn

That World Cups are not merely played, but wrung,

And half of glory is the nerve to burn.

Read it and notice your own reaction, which I think is as good a Rorschach test as any for where you stand on AI. If it still amazes you, you are probably someone who finds this technology genuinely remarkable. If all you see is a statistical engine assembling the next token, you’re probably not a fan. While the poem did not shock me the way the first one did in 2022, the VAR reference made me laugh. And a poem is now child’s play next to what these models can do.

But I think that’s the point. Effortlessly mimicking the style of a famous author about any topic imaginable is one thing; the more aware we become of what AI cannot do, the better chance we have of designing classrooms around the human qualities that make them come alive.

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

Getting AI “right” in schools has turned out to be much, much harder than most people anticipated - it doesn’t help that what “right” even means is a huge part of the problem. I have lived through several technology transitions in thirty years of teaching - the internet in the late 1990s, social media and smartphones in the late aughts, the Google ecosystem in the 2010s, and, most recently, the pandemic-era pivot to distance learning coupled with the proliferation of online learning platforms over the past five years. None of them produced anything close to the current level of polarization around AI. The poles in the public debate surrounding it (ban / embrace) often impede the likelihood of having fruitful conversations in schools. The most incendiary pieces on both sides, and the inevitable rounds of hype and dismissal, frequently crowd out reasonable attempts at nuance. As a result, most educators have found themselves caught in a crossfire they are remarkably ill-equipped to navigate. It’s no wonder academic leaders are so paralyzed and confused.

Which is why a recent letter from the President of the University of Chicago caught my attention. The most thoughtful passages read as if at least one institution is finally getting the AI moment right, or as right as the moment allows. I found myself wondering if there were any lessons other schools might learn from their approach.

The Power of Three Simple Words

On June 2, President Paul Alivisatos announced that the University of Chicago has partnered with Anthropic to roll out Claude Enterprise to faculty and staff in July, and to students before fall. The partnership headline got much of the attention (and blowback), but the letter itself is worth parsing carefully.

Alivisatos asks the community to be “skeptical, ethical, and ambitious when it comes to AI.” Those three words rarely appear together and almost certainly weren’t the result of an AI “rule of three” application. Each was carefully chosen.

The first is a reflexive stance to resist hype. The second recognizes that AI’s impact extends far beyond its immediate classroom uses. The third grants permission to use the technology in service of the school’s mission. Taken together, the triad offers a framework more schools could emulate. It cautiously captures the genuine excitement around the technology without surrendering to it, and it acknowledges the legitimate concerns without retreating into prohibition - a position shown to be untenable over the past three years.

What the Letter Actually Does

The letter does more than just take a clear stance, though it does that effectively. It also commits to a set of practical decisions. The newest versions of the tools will be made available with university data protection which means employees currently using private accounts no longer need to. Extending Enterprise to all employees - faculty, staff, and operations - rather than limiting it to academic use is a structural commitment few institutions have made publicly, an implicit recognition that frontier-tier access matters across university infrastructure, not just in the classroom.1

Crucially, the letter grounds AI engagement in what Alivisatos calls the “habits of mind and standards of judgment that define this university.” AI integration is positioned not as a departure from UChicago’s existing intellectual mission but as a continuation of it. The letter also explicitly rejects the “AI is always better” framing: “In some cases, it is clear that AI tools can be brought to bear with the desired results. In others, we will do better by retaining our previous approaches or even developing new policies to expressly protect against misuse.” This line should assure critics who fear AI is going to infuse every aspect of the student learning experience.

Speaking directly to students, Alivisatos writes that:

… the University has a duty of care to ensure that the education offered to you is responsive to these technological developments by teaching you how to think with machines, how to think without them, and how to think about them. (emphasis added).

That single sentence contains more practical wisdom than the densest AI policy memos.

Faculty resources linked in the letter are home-grown, curated, specific, and authentically the product of sustained faculty work. They represent the kind of investment of faculty time across multiple departments that does not materialize in a few months. Two new faculty committees - one on AI and academic life, chaired by William Hubbard of Law and Economics, and an advisory committee on AI and education, chaired by Emily Lynn Osborn of History - signal that institutional commitment is ongoing, not a one-off announcement.

The Process

Chicago’s position didn’t appear overnight. It is the visible output of a faculty process that has been running in the open since early 2025, when Provost Katherine Baicker convened an AI and Education Working Group. That group’s report gave the University its governing phrase: be skeptical, ethical, and ambitious. I haven’t been able to read the report itself (it sits behind a university login), but Alivisatos has described its thrust the same way in every public message I can find, from an October note to faculty through this letter.

The June announcement is the logical next step and culmination of that process. UChicago has run its own tool, PhoenixAI, a custom platform built on OpenAI’s models, available to faculty and students for over a year. Its Becker Friedman Institute has put Anthropic’s models in the hands of its economists. None of this was improvised. Most institutions cannot point to the same trajectory.

A Year Ago - CSU’s Announcement

The UChicago letter contrasts nicely with what happened at California State University last spring. A year ago, the CSU system signed the largest single-institution ChatGPT deployment in the world, a $17 million, 18-month deal with OpenAI rolled out across 22 campuses. CSU renewed in May 2026 for $13 million annually over three years. The New York Times Magazine recently characterized the result as “chaos.”

The UChicago decision, by contrast, emerged from a process the community could see coming. While clearly there is some disagreement within the UChicago community2 - no decision in the current climate around AI is immune - my admiration stems less from the decision3 itself than from the principles and process that produced it. The financial terms of the deal have not been disclosed, which is a significant criticism the partnership has invited. The framework, however, is public. A school community deserves open, defensible principles - and on that score, UChicago has demonstrated intellectual seriousness that suggests they can change course if the rollout ends up doing more harm than good.

No Easy Cases

When we study the Supreme Court in my Government and Politics class, I tell my students that no case that makes it to the Supreme Court is “easy.” Cases that survive the appeals process and are granted certiorari pose hard legal questions. If the decision were obvious, it wouldn’t be in front of the highest court.

The same logic applies to school leadership decisions on AI. One reason school leaders have been reluctant to take a strong stand is that the landscape is incredibly fraught. The technology is improving faster than anyone can follow, threatens the essence of what schools have been doing for decades, and allows students to complete work with little to no evidence of learning. It is forcing a logistical, pedagogical, and existential crisis on a scale schools have rarely had to handle, and it is not going away. At the same time, as Alivisatos notes, “faculty and researchers at all levels are using AI systems in astonishing and inspiring ways.” The juxtaposition of these two realities is what makes determining where AI might have a place on college campuses so difficult.

Alivisatos’s decision will appeal to some and upset others. That comes with the territory of being a school leader. Many schools will not arrive at the same place as UChicago. The relevant question is whether the decision was reached through reasoning that critics can actually engage with. In this case, it was. Critics need to specify exactly what was wrong with the reasoning, not just object to the outcome. Alivisatos’s language is deliberative and forward-looking at the same time, which is precisely the posture schools need to be in right now.

What Other Schools Can Learn From UChicago

Schools downstream of higher education - especially the K-12 institutions I work with directly - now have at least one model of what thoughtful engagement with the AI question looks like. Schools need principles, not just policies. Policies are rules; principles are how those rules adapt.

At the end of the day, any letter to the community about its AI principles is just words. But words matter when they are coupled with significant institutional decisions. The choice to partner with Anthropic (or OpenAI or Google, for that matter) is a major commitment. It’s incumbent on any school, not just UChicago, to follow up on its words with ongoing adjustments, calibrations, and actions across years, regardless of which direction they decide to move in. Many schools still have not chosen to do this kind of work. The few that do are starting to draft Phase II of the AI conversation in education.

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

One of the most unique things about teaching is that we exist within discrete units of time. Our year is the academic calendar, not the seasonal one, and each carries its own distinct themes, accomplishments, and struggles. 2025–2026 was a strange one for me – in some ways one of the more difficult of my thirty-one years, in others one of the most rewarding. I wrote a couple of weeks ago about a kind of AI exhaustion that had settled over me. It still lingers.

Still, maybe because I can practically reach out and taste summer right now, with the daily grind of classes receding, I'm already feeling the pull to return to the AI question in earnest (that didn't take long!). This summer I plan to approach it from three angles: the practical, the pedagogical, and, increasingly, the philosophical.1

The Practical: What’s Hype and What’s Real?

By practical I mean keeping up – not only with the models and the updates, though certainly those - but with the larger issues lurking in the background. I follow the news closely; I can’t help it, it’s my nature. During the school year it’s easy to get lost in the latest shiny thing, but summer is the chance to step back, survey the landscape, and also to play. Playing with AI has only gotten more interesting as the possibilities keep expanding.

Much of the AI discourse2 continues to traffic in sweeping claims that bear little connection to my reality on the ground, fueled by the news of the day and usually folded by each commentator into whatever they already believe. Little of it is grounded in what AI actually does - much of the pontificating marinates in abstraction.3 AI is not all one thing, and yet so much of the writing about it draws confident conclusions and makes definitive predictions that either avoid nuance or refuse to engage a single counterargument. This applies to all sides.

The other end of the AI discussion occupies a more granular and niche slice of the conversation - prompting guides, productivity hacks, and inside baseball secret AI strategies offering to give users an edge in … something. Many of these posts fixate on models, examples, jargon, and workflows. This kind of writing is not informative beyond its narrow purpose and simply assumes AI is mostly beneficial. The loudest of the AI hype crowd online are often these writers and their followers.

The two camps rarely talk: the skeptics don’t engage the boosters, and the boosters write off the skeptics as out of touch. Almost everyone is talking past each other. This has more or less been the case for the better part of two years.

My third way has been to simply move forward using my own experience as the best test case. I’ve seen the studies and try to read all viewpoints. I have some strong opinions but I reserve the right to change my mind, which I’ve done on more than one occasion. I have a list of tools I want to try this summer, mostly to see for myself what’s overhyped and what isn’t. I’ve had success building some simple static sites this year, but I know I’ve barely scratched the surface and I want to explore more.

This is a hobby, but AI has allowed me to do things I’ve wanted to do for years: my current goal is to architect full-fledged online academic courses based entirely on my own specifications, the kind of project that would not have been possible a few years ago. I’ve already started a total redesign of my Government and Politics course I’ll be teaching in the fall, inserting House and Senate trackers for the midterms, a navigable glossary of political terminology, and all eight major units embedded on a single site. I’m unsure whether any of it will justify the hours I’m going to invest, but I’ve never found a better way to learn something than to just do it. The beauty of AI is that if it’s not working or you don’t understand something, you can ask.

It’s also why I think a lot of the negative coverage around AI and education right now is aimed at the wrong target. Much of the commentary is still fixated on the student writing problem – who is using AI, how to catch it, or whether to allow it at all. That concern is legitimate, but I think it’s trailing the reality of where things are headed. Students outsourcing their writing won’t stop being a problem, and in many ways it’s already changed many students’ behavior. But using AI as a default search engine or a writing companion is not where the power is going to lie, and my hunch is that a year from now the use cases that matter will look more sophisticated than the ones we’re arguing about today. I’ll be curious if we’re still litigating the essay with the same intensity twelve months from now.

Pedagogy: Where Does AI Actually Fit?

This is the critical question that matters most for educators. There’s no question AI went through a negative cycle this year – backlashes to backlashes and visceral reactions to screens in general, especially as AI got more powerful and intertwined with everyday life. That makes the environment for talking about it in schools more fraught than ever, with contradictory and often incendiary information competing for attention at exactly the moment when the decision-makers need more clarity, not less.

In a year of conversations with school leaders, a pattern has emerged. More of them realize they need to take some kind of principled stance beyond just policies, because it’s clear that the issue is not going to disappear. But centering a school’s pedagogical questions on AI in particular, I’m increasingly convinced, is the wrong strategy. The schools that thrive moving forward will be the ones with a clear sense of mission and the flexibility to meet the pedagogical issues raised by AI as they continue to evolve. K-12 schools that go all in right now without careful planning, as well as schools that refuse to engage at all, will find themselves either reinventing the wheel or constantly playing catch-up.

So what does thoughtful discussion look like? I’m more and more of the view that it mostly looks towards practices that worked before ChatGPT launched – prioritizing student engagement, attention and time on task, scaffolded learning - the same things skilled teachers have done for decades. Teaching in the age of AI does not require a new kind of teaching - at least, inside secondary classrooms. Anyone who has stayed in this profession long enough is used to adapting to new environments. These techniques can be done with or without technology. Last August I wrote about my AI-Aware Strategy for the Year Ahead, and the honest truth is that centering AI as the main driver in the conversation was the wrong move. What I want this fall is simpler – not an AI-Aware Strategy, but simply an Effective Teaching Strategy, where AI just happens to be something we all have access to.

The Homework Question

The real AI disruption over the past two years is the impact it’s had on the work we assign outside the classroom, and that’s where serious rethinking needs to happen. Many teachers have already started. The uncomfortable truth is that the focus on “learning” – which should have been the priority all along – has drifted further and further from daily practice. That sounds counterintuitive, but for a lot of reasons we stopped making actual understanding the main goal of school a long time ago, and AI is now forcing that conversation at a higher level than we’ve had it in years. That should be a good thing in the long run.

A real and growing number of students are already functionally dependent on AI for much of their written work, and that is only going to grow. A second, likely much smaller cohort, will continue to deliberately avoid it – these are the ones who fell in love with reading and with the slow work of writing, either on their own or under the tutelage of an inspiring teacher. Ideally, we can maintain and even increase the size of this group in the coming years.

But the largest group by far is the middle: students who haven’t thought much about any of this, don’t especially care about the AI-and-writing angst, and are mostly trying to get through the day as best they can. They reach for AI in some moments and not others and don’t think very hard about it either way. In every one of these cases, AI has changed the calculus for how students approach a given task, and no AI detector is going to fix it. This is the group that needs the most attention.

None of that means we should stop asking students to do work outside of class. It does mean that any of that work, if we intend to assess it in a high-stakes way, carries an authenticity problem that is not going away. Harsh penalties may deter the worst offenders, but given the infinite number of ways a student can use AI in the context of an assignment, committing finite resources – in time, energy, and trust – primarily to enforcement feels like a losing proposition in the long run.

This is also why AI is both the cause of and a potential solution to the problem of out-of-class work – the cause, because students will keep outsourcing their thinking if we don’t re-examine the tasks we give them; a possible solution, because we can use AI to help us rethink those tasks. I wrote about this previously and I still believe it: in the absence of access to an expert curricular designer, AI can be used to help a thoughtful teacher analyze what may be working in their course and what isn’t. Running your own curriculum through that kind of audit is a solid place to start, and something I’ll do more of this summer.

That’s for the near-term. A more hopeful future, perhaps hard to envision in the current climate, is one where students come to see AI as something that can augment their own abilities rather than replace them, and where school teaches the core skills first, so that the tools are actually useful once students reach for them. I’m not sure how we get there, but as I watch another group of seniors head off to college, I want to believe that we can.

Student Saviors

What got me through the year, in the end, was the reason I’ve remained in the classroom for over three decades – the kids. Time and time again, my students saved me from myself. Real, genuine human connection is the heart and soul of any classroom. We learned this instinctively during the pandemic. Without hearing the enthusiasm in a student’s question, their frustration when they still don’t understand, and their joy when something actually clicks, none of this really matters. No chatbot can read those cues (despite what the new voice models are claiming). Even if they could, it’s clear that’s not what teenagers want.

My students this year were curious, hardworking, and genuinely wanted to learn. They asked tough questions and put in extraordinary effort, and it is clearer to me than ever that what they want most is a caring adult they respect at the front of the room. They prefer the kind of direct attention that only we can give them. My AI Club lunches, Saturday debate tournaments, and Ethics Bowl run to the finals were all highlights of my school year.

Where AI can help is in the hours we aren’t there – when students are guided and motivated to use it to deepen their understanding, apply new research skills, or get feedback on a topic beyond their teacher’s expertise, ideally after they’ve built the skills that let them use it well. The tyranny of assessment hovers over all of this, distorting and too often undermining the very learning it is meant to measure. In an ideal world we help students reach their full potential with these tools if they want them, once that foundation is in place. This is, of course, easier said than done.

What I don't think serves anyone is pretending AI hasn't already worked its way into so much of student life. Pretending the situation doesn’t exist, or trying to ban and/or punish our way out of it, only guarantees that students will use AI in the shadows. My message in the K-12 setting is to center the learning first, foregrounding student needs with the understanding that many of them are already turning to AI. A former colleague used to say that no technology tool was inherently good or bad - user intent coupled with the potential for positive learning outcomes made all the difference. It would be a major breakthrough if we reach the stage of discussing AI in school settings where those considerations drive the conversation.

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

Last Tuesday, Google held its annual I/O conference, and I’m one of those people who watch the entire keynote. I came away both overwhelmed and underwhelmed. The new models are genuinely impressive, with real agentic capability and a new fast-and-cheap model, Gemini 3.5 Flash, that does frontier-level work at a fraction of the cost. Charts sped by showing AI usage exploding across the Google ecosystem. The technology is real, and it is improving ahead of schedule. What frustrated me was not the technology. It was the way Google chose to show us what they want us to use it for.

A handful of lines also stood out, particularly from Sundar Pichai, Google’s CEO, and Demis Hassabis, the CEO of Google DeepMind and the company’s resident visionary. Pichai casually folded in two examples of how people are using AI: students studying for final exams, and his own use of it to “make sense of my parents’ doctor visits.” Later, Hassabis stated, essentially as fact, that we are just a “few years away” from AGI. I know these conferences are victory laps, with their obligatory pauses for applause, like a State of the Union address. Even so, Pichai reached for two of the most contested uses of AI in the culture right now - student work and medical diagnosis - and presented both as if the debate were over and decided in their favor. Hassabis has moved his AGI timeline up to the point where he’ll say it flatly, on a stage he knows the world is watching. Each of these lines struck me wrong. But it was the demos where they really lost me.

A year ago I wrote about a significant development most people missed that rolled out at the same conference, and we’ve all been living with the consequences ever since. AI Mode, built directly into the search bar, is now just part of the Google experience. This made AI unavoidable for the billions of people who use Google search, including our students. This year, Google is clearly doubling down on its AI infrastructure investment. To what end?

All That Intelligence, for This?

The demos, one after another, were mostly small in a way that’s oddly hard to describe. Gemini Spark, the company’s shiny new personal agent, was put to work organizing a block party. In Josh Woodward’s demo it tracked down RSVPs, updated a Google Sheet, drafted emails to the neighbors who hadn’t replied, and built a hype deck in Slides, all of it running in the background. A second presenter, Robby Stein, showed something different: generative UI in Search, where it built a custom app on the fly, in this case a weekend planner for his family, which he then shared with his wife for her seal of approval. A little later came the universal shopping cart that follows you across the web, and that was the clincher for me. All that money, all that intelligence, and it was mostly pitched as a better way to throw a barbecue and a faster way to check out after an online shopping spree.

The promise of all this is that AI does things for you. The demos kept showing me how AI could build things, even directly within the search box.1 I laughed at the weekend planner, because I know exactly what would happen if I built an app to plan our weekend and sent it to my wife. She would probably throw her phone at me. Who actually wants this? When I use Google Maps to find a coffee shop, I want the coffee shop, not to spend time fiddling with a custom-built coffee-shop-finding dashboard I can come back to later.

Over and over, they handed me solutions to problems I don’t have. So much of it felt less like help and more like homework, an endless proliferation of apps and dashboards for things families have managed on their own for decades. Google’s bet is that making these will become so fast and so frictionless that we’ll reach for them without thinking. Maybe they’re right. I’m skeptical.2

It’s Shopping All the Way Down

The part that left me most cynical is how obviously Google wants us to shop more. People already shop across Google more than a billion times a day, and the company intends to make that even more effortless. The new shopping agents work around the clock. Your cart follows you everywhere, hunts for deals while you sleep, and can be authorized to buy on your behalf with you out of the loop. One slide called it “shopping with superpowers.” The commerce segment ended on the words “happy shopping.” Then, just in case the message hadn’t landed, they scored the transition to Depeche Mode’s “Just Can’t Get Enough,” hitting Gen X squarely in the solar plexus.

I’ve no doubt some people may love this. Still, a discount on something you’d never have bought without a shopping agent doesn’t count as “savings.” It’s just a more efficient way to spend more. “Shopping with superpowers” is really a demand that we all shop with superhuman discipline, which is not a thing most of us are good at. We’ve already built a world where we can want something at eleven at night and have it on the doorstep the next day. That last layer of friction, between the impulse and actually clicking Buy, was arguably the one thing protecting some of us from ourselves, and Google’s plan is to remove even that. In a spring when gas is closing in on five dollars a gallon, inflation is climbing, and consumer confidence just hit an all-time low, do young people, and the rest of us for that matter, really need to be trained to turn every passing whim into a transaction?

Glasses That Can Whisper the Answers

Near the end came the demo I’d been dreading, and the one that should scare teachers most. Google’s first audio glasses, which speak information privately into your ear, arrive this fall; the display glasses, which show it on a lens in front of your eye, are still in limited testing. The demo was kept deliberately mundane, because the general creepiness of AI headgear is the part Google would rather not dwell on. Shahram Izadi, Google’s XR lead, and Nishtha Bhatia walked through ordering a cold brew through the frames, summarizing muted texts, and turning the crowd into a cartoon on her watch. Cool, I suppose, but do consumers really need this?

For a teacher, a device built to feed you information invisibly, whispered in your ear or shown on a lens only you can see, is an assessment-integrity problem on a scale we haven't faced, and it is not waiting for Google's fall launch. You can already buy prescription glasses with AI built in. Meta has sold millions, and a student in my class had a pair last year. Can we ban prescription AI glasses? The blue book and the in-class exam, the defenses we've fallen back on this year, do not survive glasses that give students answers in private. Beyond cheating, the glasses are just the next layer of AI envelopment Google keeps working on to make it all feel inevitable, arriving at a moment when the backlash is real and a vocal public is asking everyone to slow down. The whole keynote felt tone-deaf to it.

The Argument They Skipped

The debate about AI and education is the one the keynote treated as a closed case. The references that did appear were all upbeat - Pichai's line about students prepping for final exams, and Josh Woodward's quick run through Gemini's student features, guided learning, practice tests, and a nod to NotebookLM.3 The closest the keynote came to demonstrating actual learning was Robby Stein's astrophysics moment, where Search built a custom interactive visual explaining how black holes affect space-time. Even that was staged as an exciting feature of what students can do. Yet the question that keeps teachers up at night, whether these tools are actually helping students learn, never came up. Google ignored the most contested issue in my world and declared it settled while the rest of us are still figuring it out. It won the argument by declining to have it.

After the keynote, Dan Fitzpatrick (no relation), a highly sought-after voice in the AI-and-education world, sat down with Chris Phillips, the man who runs education at Google, and came away with a much more nuanced, teacher-centered case the keynote never bothered to make. His piece is worth reading in full.

Two of his takeaways landed, for different reasons. The first described one of the most impressive demos in which Antigravity 2.0, Google's revamped agentic coding platform, built a working operating system from scratch, mostly on its own, for under a thousand dollars. Building something that complex for that little has real ramifications for developers, but it is not likely to impact teachers any time soon. It is just another example of the gap between what the frontier can do and what the 2026 classroom is actually concerned with.

His larger point is that Gemini 3.5 Flash is now good enough to do almost everything schools are asking for, and the new price point will lower the cost barrier keeping AI out of districts willing to make the switch but unable to commit the resources. I suspect it will make AI in schools more likely, not less.

Though Dan’s piece might be interpreted as more bullish than I am on where AI in schools is headed, it did not offset my visceral and mostly negative reaction to the keynote itself.

Even though it’s clear from his reporting that Google has a thoughtful account of what its tools do for learning, it’s telling that it chose not to put that on the stage.

The Promise vs. the Product

The keynote opened on a sizzle reel of possibility - a young boy on a tablet doing something undefined but clearly interesting with AI, a teacher mid-music-lesson, a few small-business owners, over a voiceover about solving disease and “making something that matters.” There were legitimate use cases: a Korean immigrant who used these tools to build a hiring platform for her neighbors, and creative AI tools for musicians and filmmakers.4 It closed on Hassabis invoking the “foothills of the singularity.”

Hassabis is the AI paradox in one man. He came out early to announce that AGI is just a few years away, and he returned at the very end as the closer, touting AlphaFold, Isomorphic Labs chasing the goal of one day “solving all disease,” and WeatherNext, the forecasting model that called Hurricane Melissa’s path days early, in time to get people out of the way. These are all real ways AI has advanced scientific development. He promises a “new golden age of scientific discovery” that improves lives everywhere. The demo that ran just before that close was the cold brew, ordered through a pair of glasses. From a cold brew to the singularity in a single cut. Everything between his two appearances, the apps, the shopping, the glasses, was the version of AI its skeptics fear most. Which future is Google actually building?

To What End?

Pichai mentioned that Google expects to spend something like $180 to $190 billion on infrastructure this year, roughly six times what it spent in 2022. Spark, the advanced agentic tool covered in the keynote that works while your laptop is closed, requires a Google AI Ultra plan, which runs between $100 and $200 a month. What's free for everyone is the new search box, the on-the-fly generative results, and the cart. Google would tell you the cart is what pays for the moonshot. Maybe. AGI is held out as a tantalizing promise, but the cart is what turns a billion shopping searches a day into the money that pays for it.

There were genuine standouts - the aforementioned Antigravity build and Gemini Omni, the new video model, were both impressive. With Google Flow’s editing tools I could see both the fear and the possibility in it for anyone who fools around with photos, film, and music. We will likely absorb Spark, the in-search app builds, and the universal cart the same way we absorbed AI Overviews, without too much fanfare. Maybe I’m just grumpy, but a lot of it felt more like gimmicks than upgrades.

This all tells me that with agentic AI now baked into how we use Google, the fantasy of “banning” AI in schools should be over, if it was ever a serious position to begin with. Yet we still barely know whether, or how, to teach with these tools at all. For the first time, as someone usually curious and fascinated by this stuff, I’m feeling more AI fatigue than excitement. Maybe I am just suffering from AI exhaustion.

Still, what stuck with me most was the shopping. The direction Google is pointing, as best I can tell, is to get me to buy more, build disposable apps to help me do it, and surrender to a world where every experience is something to digitize, optimize, and monetize. AGI promised something bigger than this. Do I really need a customized shopping experience, or bespoke apps for using the web? And do I want a pair of glasses talking into my ear or digitally interpreting my world so I never have to engage with it? The 2026 keynote was a tale of two futures, and the one that feels likeliest after watching it is the one I want least.

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

It’s mid-May and there is light at the end of the tunnel. Our school year ends in three weeks. Instead of anticipation, what I feel is something closer to exhaustion. The exhaustion is partially due to an especially brutal workload, but it also comes from spending nine months watching the gap between what AI actually is and how schools engage with it grow wider, not narrower. I feel like the conversation itself is stuck.

Below I’ll share a few observations as we wind down the academic calendar. None are especially surprising or groundbreaking. Taken together, though, they suggest significantly more work ahead for teachers, students, and institutions. My key takeaway is that the 2025 - 2026 school year will be remembered as the one where AI became ambient, unavoidable, and still mostly unaddressed.

Ambient AI

At the risk of stating the obvious, AI is everywhere. It’s still mostly associated with ChatGPT, but as of May 2026, unless you live under a rock, almost everyone is aware that several major companies have powerful AI models. More importantly, regardless of whether you know the names, understand the difference between Claude Code and Codex, or have experimented with any agentic capabilities - everyone on the internet has encountered AI images, audio, and sophisticated AI-generated content.

This was not true a year ago. The deep reasoning models had only emerged at the end of 2024 and I recall shocking teachers with demonstrations of what they could do last spring. The speed with which add-ons, plug-ins, and upgrades have rolled out means we are well past the ability to shock.

The implication for educators is simple. Unless we are prepared to keep our students from accessing Google, they will always be able to use a powerful frontier AI model in the form of Gemini through AI mode. I’ve argued since last June that teachers who claim they “forbid” AI are engaging in semantic theatre once a student is on the internet - we all encounter AI overviews whether we like it or not. Google’s AI mode allows for the critical cognitive off-loading educators fear, with the ability to have an LLM write, analyze, and work with uploaded documents baked into search. Literally the same capabilities students can access through ChatGPT. They do not need paid models.

Any notion that AI will not have a significant impact on teaching and learning is pure nostalgia.1

Schools Are Finally Reacting, But The Gap Is Still Growing

The good news is that schools are finally taking AI seriously beyond writing policies. More administrators and district leaders are grappling with what’s actually happening rather than merely drafting documents that are increasingly opaque, difficult to enforce, and not designed to withstand the speed with which the technology is moving.

The bad news is there is no consensus on what the reaction should be. The recent trend is to zero in on devices themselves - a movement grounded in legitimate concerns but one that also conveniently feels like it “solves” the AI problem. If students don’t have their phones or access to laptops, there’s no technology to turn to. Essentially, it asks us to return to an idealized past where kids all wrote by hand and played outside. As a father of a 6th grader, I see the appeal. I really do.

But the majority of school districts already have technology so embedded in their infrastructure that an unraveling is not nearly as easy as it looks. Many are going in the opposite direction, having invested heavily in a future where AI is a significant part of their mission. With over 13,000 districts in the U.S., we will see almost every permutation along the technology and AI adoption spectrum. At least these experiments will give us more data and more opportunities to evaluate what works.2

The deeper problem is that even the conversations finally taking place are still mostly happening in the context of 2023 AI. Few schools that I know of are seriously engaging with where the models are now and where they are heading. Over the past year, AI has moved well beyond the text generation that prompted most school policies - it now builds functional applications, conducts multi-step research, and produces professional-quality documents and presentations, capabilities that most school conversations haven’t begun to address. And these are the paid frontier models - what the average teacher or administrator is actually encountering is likely already a generation behind.

I agree with those who argue that mastering current models shouldn’t be the goal of AI literacy - they will continue to evolve and what we use in May 2026 will not be the same as September 2026, let alone May 2030. But I depart from those who claim that learning AI is “easy” and that our responsibilities as educators do not include preparing students for the world they are actually entering.

The reason I disagree is the same reason I would have pushed back in the early 2000s when internet access arrived at scale. We had responsibilities then to demonstrate best practices, explain how to evaluate information, and model effective use. Some of those early digital literacy strategies didn’t hold up over time - and that will likely be the case for some of what we teach about AI today. But talking about AI in the abstract without showing students what these tools do, how they work, and what kinds of output they produce is not serving them well. Especially when we know almost all of them are using it outside the classroom.

The Big Questions Are Finally Landing

Perhaps the most meaningful development of the year is that AI has forced educators to confront questions about purpose. More people - even those most viscerally resistant to AI - are recognizing that these systems will likely be a significant part of our lives going forward. Skeptics are also wrestling with something harder to dismiss: that AI can organize, synthesize, and generate information in ways that people engaged in knowledge work find genuinely useful whether or not those opposed to it use AI themselves.

These capabilities are forcing hard, but necessary conversations. Not simply about cheating, which dominated the past three years, but about purpose. What are schools actually for when obtaining and presenting information - in virtually any format we require - is approaching zero cost? What does a research assignment mean when AI can search, synthesize, and organize information faster than the student writing it? These are existential questions, no longer the province of the futurists and technologists who have been raising them for years. They are finally landing in faculty meetings and department conversations - perhaps under duress, but landing nonetheless.

Maybe it was never about AI in the first place. It’s always been about the value of what we assign and what we ask students to do - questions we could avoid as long as the products students handed in were plausibly their own. It’s dawning on most of us, if we hadn’t realized it already, that the skills hardest to automate are the ones the humanities have always tried to teach: empathy, curiosity, the ability to ask good questions, and the patience to sit with ones that don't have clear answers. It's worth noting that many in the industry are arriving at the same conclusion.3

What It Cost Me This Year

I want to add one observation that is entirely anecdotal but may resonate with other teachers. I covered less content this year. The constant vigilance about AI, the reduced number of assignments I gave outside of class, the time spent on research projects completed under supervision rather than at home - it all added up. I simply could not cover as much material as I have in previous years.

I don’t know if this was specific to my circumstances or something more widespread.4 But it strikes me as one of the hidden costs of this transition that we are not yet accounting for. AI has added more monitoring, more in-class assessment, and more rethinking of assignments to a curriculum that was never conceived with it in mind. If schools don't change their approach, something has to give - and right now, for me, it's content. Maybe that’s not the worst thing when virtually unlimited access to information and the ability for a motivated student to learn on their own are higher than they’ve ever been. But Sal Khan himself recently acknowledged that Khanmigo was a 'non-event' for most students - they simply didn't use it.5 Students are constantly using AI, just not the way adults design it for them.

As I Head Into Summer …

2025 - 2026 is ending more or less where we started, but with even more uncertainty - and significantly more work ahead within our schools, with our students, our parents, and our colleagues, to catch up to the reality of what AI has done to the academic project. I can’t imagine 2026 - 2027 won’t bring more of the same.

I’m exhausted by it and I suspect many teachers are as well. But the questions won’t disappear just because we’re afraid of asking them. If anything, we’re just getting started.

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

In March of 2020, millions of teachers and students across the country pivoted online for distance learning because of the pandemic. Districts that did not have 1:1 laptop programs swiftly invested in Chromebooks for every child. Despite the imperfect nature of the experience, we would not have been able to teach or connect with our students without those devices. Chromebooks became a lifeline between many students and school, particularly those in the least resourced communities. Yet, even for schools with 1:1 devices before 2020, most never carefully designed pedagogy around providing every child with one. Recent cell phone bans and the rise of generative AI have created a perfect storm for an anti-screen movement. The impulse is both understandable and rooted in legitimate concern. My fear is that the same lack of forethought that led to bringing laptops into the classroom in the first place will result in an overreactive push to eliminate them at exactly the moment students most need to learn how to navigate an AI-infused digital landscape.

Recent mainstream media articles report an accelerating trend.1 Parents and policy makers are demanding technology rollbacks in Los Angeles, New York, and elsewhere. Some middle schools have limited student access to Chromebooks. Others are trying to get rid of them altogether. The reporting describes negative reactions to the prevalence of screens in classrooms, with AI as a visceral catalyst that has helped transform long-standing discomfort into a national movement. The backlash is both understandable and real, and the responses range widely - from sensible restrictions on where, how, and for whom the devices are used, to outright calls for removal.

Bans are not the same thing as reforms - the distinction matters.

Any teacher who has spent time in a classroom over the past decade knows that screens can be the death of attention. I am sympathetic to the occasional urge to remove them. Yet, for lessons and activities designed well, digital devices that connect to the internet are one of the most powerful tools ever devised. Everyone reading this piece is either using a cell phone, a tablet, or a computer. Internet access is a necessity to participate in modern life.

The clash of these two realities - the difficulty of maintaining attention in the presence of a phone or computer that can deliver and create unlimited content - is at the heart of the conflict. At what age and in what circumstances can students be expected to responsibly use a device? How can teachers leverage intentional digital instruction and what can schools do to best ensure those devices are used appropriately? What does it mean when every child has their own device in the current paradigm, and what would it mean if we suddenly took these away?

The fundamental question around the issue is: Are the devices inherently the problem, or is the fact that we never developed real pedagogy around them - despite over a decade of trying - the real culprit? Some in the screen-removal movement increasingly claim it’s the laptops themselves. I've seen Chromebooks and classroom technology work for kids. Like almost everything when it comes to teaching, the results are context and classroom dependent.

Chromebooks were issued almost as a wartime emergency measure in 2020 and 2021. The devices have subsequently let in, among other things, first YouTube and now AI as a Trojan Horse, causing many of the problems outlined in the articles. Rather than ditching devices entirely, a better strategy is to pinpoint where the technology is doing more harm than good. Pulling the football away now, after a decade of telling kids and teachers digital and media literacy is essential, signals one thing above all: the adults have no idea what they are doing.

What the Movement Gets Right

Most of the horror stories in the recent coverage focus on the K-8 case, where the concern is clearly warranted. Before the pandemic, only about forty percent of districts had given devices to every elementary student. By March 2021, that number had jumped to eighty-four percent.2 The pandemic's distinctive contribution was pushing screens down to the youngest students at scale, before anyone had figured out what those devices were supposed to be doing in elementary classrooms. Putting screens in the hands of seven and eight year olds as a primary means of learning is not pedagogically defensible (though I suspect most elementary school teachers would chafe at the characterization of what they are doing). Nevertheless, the bombshells in the reporting - a seventh-grader in Wichita accessing more than 13,000 YouTube videos over three months on his school-issued account, and a Los Angeles 4th grade classroom where an AI art assignment produced sexualized images of a children's book character - are damning evidence that issuing 1:1 devices to younger students is walking a tightrope at best. It doesn’t take a neuroscientist to support eliminating or, at the very least, significantly restricting screen time for kids still learning how to read.

The Larger Failure We’ve Yet to Grapple With

The stories about Chromebook misuse and unlimited access to YouTube treat the screens themselves as the problem. A more nuanced view is that most districts that issued devices never built the institutional capacity to manage them. Many schools that gave kids the keys to the kingdom in 2020 did not plan for what classroom use would look like after the emergency. For example, as the WSJ reported, internal Google documents showed that by 2019 the company itself acknowledged the YouTube experience in schools was broken and that filters were trivially easy for students to bypass. Schools have known this for years. Most have done very little. But some have used a scalpel rather than a hammer: Watertown Public Schools in Massachusetts blocked YouTube districtwide in January. They did not eliminate Chromebooks.

What the Reversal Would Actually Cost

Banning devices from older students as well as younger ones also does not realistically engage with what a full reversal would require. More than 88% of public schools now operate 1:1 device programs, according to government survey data cited by the Wall Street Journal. Schools have rebuilt their pedagogy around digital delivery for more than a decade. Teachers have rewritten units, redesigned assessments, and built classroom workflows that assume students can access shared documents, online texts, video resources, all layered on top of learning management systems.3 Standardized assessments are now digital. For many students, a school-issued device is their only means of completing their work in the upper grades. The infrastructure is not removable on a whim. It is replaceable, but not in a year, and not without enormous transitional cost. We rolled these devices out under emergency conditions in 2020 because there was no time to plan. Even though we have more time now, nothing suggests we are using it any better. Recent research on cell phone bans in the United States and on Australia's under-16 social media ban shows how hard it is for bans to deliver the outcomes their advocates promise.4

The K-8 Case for a Ban Is Not the 9-12 Case

The screen-removal movement is least defensible where it shifts from real concerns about K-8 into a blanket case for total bans. The districts at the forefront of the movement are pulling or scaling back 1:1 for elementary, and a few are now extending that to middle school. McPherson Middle School in Kansas collected back student Chromebooks in December and now keeps them in classroom carts, available for specific assignments. Bend, Oregon blocked YouTube for lower grades and is reviewing high school access where it remains the most-used app. Granville County, North Carolina is phasing out 1:1 for elementary while keeping it for upper grades. Even the most aggressive recent moves are almost all grade-specific.5 The rhetoric, however, is not making the developmental distinction explicit. More significantly, if we overcorrect and ban all devices without thinking through the consequences, re-introducing them later will be that much harder once the political moment passes. Better to do the slower work of assessing what we have than to make rash decisions that cannot easily be undone.

Chromebooks for second graders are entirely different from high school seniors learning to research with AI in preparation for college. Younger students still primarily need handwriting practice and tactile, deliberate work with physical tools in the classroom. Twelfth graders are about to enter an environment in which AI is already embedded across every common platform, and most of their academic experience will be mediated through devices. These are not the same question. A screen-removal movement that treats them as one risks landing somewhere that will likely need revisiting within a few years. Too wide a pendulum swing will compound the mistakes.

A More Thoughtful Path

Most of what is actually being reported and adopted is not a ban. It is a rollback. The districts pulling back are restricting YouTube, eliminating devices for the youngest students, scaling back from 1:1 to classroom carts, and developing screen time limits for higher grades. These are scaled-back approaches that require careful management but preserve the option to use the devices when it makes sense.

The same type of approach could be applied more broadly. Restrict use in lower grades. Conduct better and more rigorous research in middle and high school on device usage. Lock down YouTube and other sites of maximal distraction. Give parents real opt-outs. Tech-Free Tuesdays and Thursdays is a more elegant solution than immediate removal. The screen-use ratio should reverse over the course of an academic career - nearly 100% analog in lower grades, moving to more realistic usage by senior year of high school. As with so many debates, framing the issue as a binary - ban laptops or keep them - is what makes the conversation so contentious.

A Year Later: What’s the Message?

What message would laptop bans send to students? A year ago, the New York Times was profiling Miami-Dade for rolling out Google’s Gemini to 105,000 high school students. To my knowledge, the Times has not run a follow-up on how that experiment is going.6 Notably, the same reporter who profiled Miami-Dade's rollout, Natasha Singer, has since written two of the articles driving the backlash narrative - but not a follow-up on whether the largest AI deployment in American public schools delivered results.

In March, the Times also reported that NYC Public Schools, the largest district in the country, released formal guidance designating AI lesson planning, brainstorming, and drafting as “green” applications. Two of the nation’s largest districts, plus a growing number of others including Chicago, Denver, and Charlotte-Mecklenburg, are leaning in carefully. Now the mainstream media is running cover stories on the screen backlash, with at least one arguing for wholesale device removal. Across the country, many states are finally catching up by developing AI literacy plans while others are pursuing rollbacks and investigating outright device bans. These messages are more contradictory than complementary.

This is whiplash for teachers as much as for students. Many state legislatures, education experts, and consultants have spent the past two years building AI literacy frameworks and integration guidelines on the assumption that students would have access to devices. The screen-removal movement now wants to remove those devices. The signal to everyone is that, after fifteen years of use, we still have no consensus on what these tools are for, how to use them, or whether they belong in classrooms at all.7 Students are at the mercy of the adults in charge, and the adults in charge keep changing their minds.

How Can We Best Serve Our Students?

My fear is that education's response to AI will be to retreat rather than to engage. AI literacy taught in the abstract is not really AI literacy at all. That is not the future many states have spent the past two years building toward. That does not mean integrating AI into elementary classrooms. It does mean addressing AI head-on, which includes explaining, demonstrating, and recognizing where students are using it. That requires them to have access to devices at some stage of their academic career.

Older students, especially those in the years prior to college and university, need hands-on experience and guidance with teachers about both practical and ethical AI use. What that precisely looks like, how often and in what context, will be debated for the foreseeable future. To pull the rug out from under both students and teachers after almost a decade of training on digital content delivery would send a powerful message that we are moving backwards.

For some schools and communities, a flat out device ban in certain grades may best reflect their values. Claiming that 1:1 laptop programs have been the problem all along, however, when so many of our challenges in education pre-date the pandemic is to make the same reactive mistake over and over again. Students will encounter screens, devices, and AI outside of school regardless of what their schools decide. For students with limited home access, school may be the only place they will encounter these tools under adult guidance. Removing the devices does not remove the technology from their lives. It just shifts where, when, and how they encounter it, and which students get any meaningful instruction at all.

The K-8 screen reckoning is overdue, and if AI is the reason it finally arrived, thoughtfully considered rollbacks happening now are largely the right response. But banning screens at every grade level, in reaction to a technology that has barely had time to be understood, would throw away almost fifteen years of pedagogical infrastructure to fix a problem most schools have not yet seriously tried to manage. Students need to learn to master the machines at some point in their education. When and how to do that in the face of emerging and powerful technologies will be an enduring question. Let schools and teachers make these decisions based on their own experiences, not in response to a reactive and contradictory national conversation simultaneously demanding AI literacy while calling for the removal of the devices students would need to develop it. We need to get this right.

I am especially interested in hearing from anyone making the case for high school bans. Tell me what I am missing.

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

A slew of frontier model updates and new capabilities have recently dropped in quick succession.1 One, Claude Design, got lost in the shuffle. While playing around with it over the past week, something became clear. Now that code generation rather than text is the leading edge of AI capability, it comes with a cost. AI is getting more complicated to use. Vibe coding was a thing a year ago2 - non-coders using natural language to make AI write code for them. What’s emerging now is something new: native interfaces specifically tailored to create things - a document, a presentation, an app, a web tool. Most people are not taking advantage. The barrier to entry to really leverage what the most advanced AIs are capable of now requires thinking like a coder.3

Unfortunately, I’ve never been a coder. Most teachers reading this probably aren’t either. An entirely new vocabulary, one I've been learning mostly through newsletters geared toward entrepreneurs, is leaking into everyday conversation: APIs, tokens, context windows, agents, MCPs, harnesses, CLIs. You can’t read deeply in the AI space now without encountering hyperventilating claims about new workflows that produce all sorts of new digital artifacts and prototypes. Software engineers, marketers, and designers are rapidly absorbing these new tools into their jobs. That’s the hype. But hype doesn’t mean things aren’t actually changing. My experience is ahead of most teachers', but the wall is real - and humanities teachers will hit it hardest.

For most of my professional life I knew, in some abstract way, that everything I typed and clicked online was built by someone in a backroom who knew how to make it work. That was the deal humanities educators made with the digital age. We taught our students how to read critically, write cogently, and think deeply, all while seamlessly using digital tools - the internet, computers, and software. The people who knew the secret language of the machines handled the machines. All that happened somewhere else.

That deal is changing.

The original promise of generative AI was democratization through natural language. Anyone able to type can access expertise. That was true for the first wave of chatbots and it’s still true today. A teacher can quickly draft a rubric or a lesson plan. Someone who knows nothing about auto mechanics can teach themselves rudimentary maintenance tasks. AI quickly became an interactive YouTube tutorial. The friction between having a question and getting a coherent answer collapsed almost overnight.

But the simple chatbot is now the floor. The most powerful uses of AI sit several levels above where most people are currently working, which means we are drifting back toward a real expertise gap. To build something really useful with Claude Design - or Cowork, or Codex, or any of the platforms sprouting up everywhere - you have to think like someone who knows what a digital artifact really is, when to deploy versus refine, how to prompt at a granular level of specificity, and when to ask for code rather than revision. Hardest of all: how to recognize when the AI is offering you a design choice instead of a final product. A mistaken entry has consequences - in token use, in time spent waiting for an output, and in the rate limits you'll hit faster than you expect.

You don’t have to write any code. You do have to think the way coders think, make the kinds of decisions they make, and understand the language they use. A new threshold needs to be crossed.

Models, Apps, and Harnesses

If you want to see the issue up close, read Ethan Mollick’s latest post on GPT-5.5. Mollick is one of the most accessible writers about AI and he works hard to translate for non-technical readers. But even his current framing includes assumptions about what general readers are tracking. If you haven’t kept up, his posts are liable to be overwhelming.

He encourages us to think about AI as three interlinked concepts: models, apps, and harnesses. Models are things like Opus 4.7 and GPT-5.5. Apps are the products that let models do real work - claude.ai, chatgpt.com, Claude Code, and Codex. Harnesses are the tools the AI uses and how it is hooked up to them. That last one is probably the deal breaker for most teachers. How many educators do you think are linking APIs to GitHub? Setting up MCP servers? Are you still with me? This is where the real power resides for serious AI users, and it is currently beyond the reach and expertise of most educators.

Things are moving so fast that even the most vigilant cannot keep up. That is becoming more true by the week. A humanities teacher reading the GPT-5.5 piece is not reading about whether students can cheat on an essay. Mollick describes generating a near-PhD-quality academic paper from four prompts using Codex and an interactive 3D simulation of an evolving town across 6,000 years from a single prompt. These are now ordinary outputs from someone who knows how to integrate models, apps, and harnesses into a repeatable workflow.

Even Mollick has acknowledged this gap. In a post from late March, writing about the most powerful tools currently available, he admitted:

These tools are terrific, but they are really built for programmers. They assume you know Python and Git. Their interfaces look like a 1980s computer lab. For the 99% of knowledge workers who are not developers, these powerful AI tools are not optimized for them.

Claude Dispatch and the Power of Interfaces

Tools like Claude Design are the industry's attempt to fix this. The capability is there but what is new is the scaffolding meant to let non-technical users reach it. The catch, based on my experience, is that the new scaffolding is itself technical. The interface is better, but the conceptual demands are still out of reach for most. Maybe that will change, but right now it’s an impediment.

Hitting A Wall

Last December I appeared on the leaderboard of a brand new platform called AI Cred4, and the founder’s follow-up post name checked me as one of the surprises - a 30-year history teacher with no coding background showing up alongside engineers and consultants. In February I wrote that Claude Cowork “just worked” for me, and I meant it. Building online tools took some figuring out, but I did it. I suspect I'll get there eventually with Claude Design, but I'm finding myself out of my depth more often than I once was.

I need to reread every sentence twice. I’m having flashbacks to my days fumbling through FORTRAN and Pascal in the 80s. The skills that got me on that leaderboard last fall haven’t gone away. The problem is everything that’s piled on top of them since. Keeping up with the models and new releases is practically a full-time job.

The tantalizing thing is that I do see the value. I understand why these tools are a breakthrough and eventually worth learning. But taking the time to master them is an opportunity cost I can’t currently spend. It all makes me nervous about what comes next.

Last week I asked Claude Design to revise a one-page professional summary I’ve used for years. The draft came back elegant and the formatting impeccable. I was inserting updates from the last few years and needed to know how best to export the file. It recommended a standalone HTML file - not PDF and not Word. Then it offered an unsolicited note: “Start a new chat to save 107k tokens of context. This uses your rate limits more effectively.”

I knew what “standalone HTML” meant because of the Cowork projects I've done. I also knew that saving “107k tokens of context” was a good thing. A teacher who recently upgraded to Claude Pro to try Claude Design might not have known what either phrase meant.

It is not that learning these capabilities is hard, exactly. You can ask the AI when you get stuck and it will generally walk you through it. What the companies are doing, however, is putting the newer models out of reach for all but the most dedicated users with the resources and time necessary to devote to it. It’s understandable why companies will invest in this. It’s not yet clear why schools should.

The Eternal Skills Question

The best argument against using AI in classrooms goes like this. Teach students the eternal skills - writing, judgment, critical thinking, the ability to refine ideas - and AI fluency will follow. Kids need to know how to think before ever touching a chatbot. Concrete elements of AI instruction, like prompting, can be learned later. I tend to agree with all of this in the abstract. The strongest version of the argument is that schools exist to build durable human capacities, not to chase whatever the workforce happens to need this year.

But the eternal skills argument assumes students will not be using AI during that time. That is impossible unless we lock them in a room without technology. Some parents would willingly make that tradeoff. Most schools cannot.

It also assumes the AI students encounter at twenty-three will resemble the AI they are using at fifteen. It won’t. Students graduating this May and June will matriculate to college and eventually workplaces where the entry-level fluency expectation is already past the chatbot tier. The AI of 2034 will be unrecognizable to them, and the judgment required to use it well is not something that can be snapped on at the end of college. AI is the first tool that talks back in any meaningful sense - it argues, flatters, redirects, and sometimes deceives. Nothing in our experience tells us how long it takes to develop fluency with something like that. Withholding AI from students until they’re “ready” assumes a static technology which currently doesn’t exist.

I'm not convinced the eternal skills will be enough on their own, or at least, they are going to have to be developed alongside the reality of student use. AI literacy is on track to become a baseline professional expectation and the cost of being late is likely to compound quickly.

Class of 2032

My middle-school daughter graduates high school in 2032. Even if you accept the argument that high school is not workplace training, not exposing her to AI and what it is capable of during her high school years feels like an abdication of our educational responsibilities. We know they're using it already - every conversation I've had with actual students confirms it. Teachers cannot use AI themselves while refusing to explain it to them. The question no one has been able to answer yet is how to do that well. Can we learn this with them rather than ahead of them? The challenge is how to teach a tool we’re still figuring out ourselves.

This is uncomfortable territory because it implies that the humanities teachers leaning into the AI conversation in schools - and I am one of them - need to learn things we did not sign up to learn. Not exactly how to code, but how to adopt the mindset and understand what computer code is. We need to recognize when a tool should be used as an artifact builder rather than a chatbot and, most importantly, when it should not be used at all. Even more daunting is we may need to adapt the vocabulary of the software engineer to the vocabulary of the classroom.

The Home Stretch

As we head into the home stretch of the school year, I have been reflecting on what has changed for me over the past twelve months. “Everyone is Cheating Their Way Through College” was published on May 7th, 2025. That is old news now, and a tired take, though I do not know exactly how much has actually changed in classrooms. There may yet be a wave of success stories about blue books and oral exams. My experience has been mixed.

The real story for me is how I have moved from a purely text-based use of AI to one predicated on the ability to generate code to do things. Yes, it is partly the “agentic layer” that was promised last year and is now taking root. But the more important point is that it is going to require an entirely new set of skills for non-coders to do the kinds of things that come naturally to coders.

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. Generative AI gave us, briefly, the illusion that we could keep swimming solely in the world of words we understood. The truth is we have been living in someone else's domain all along. We just did not have to see it. Humanities teachers may have to learn to read in a language we did not train for, if we intend to keep up.

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

It's late April and I'm running on fumes. Students and teachers everywhere can see the finish line, but it's still far enough away to be more distraction than relief. Overextended and buried under a to-do list a mile long, if ever there was a moment when a time-saving tool would be critical, this is it. And yet AI is not that tool - at least not in the way it’s frequently pitched.

The availability of powerful generative AI makes it very easy to do things that probably should not be done at all. A core skill in the coming years will be determining when, why, and for what purpose AI should be used. Because, make no mistake, it will continue to be used by millions of people, and many of those people will be students and teachers.

Six Weeks of Time Savings? Really?

If there is one message that unites nearly every report, conference keynote, and vendor claim about AI in education, it is that AI can save teachers time. The January Brookings report on AI and students summarized the consensus:

AI’s benefits extend to teachers: by reducing time on tasks, AI allows teachers to focus on individualized student attention and enhance curriculum and instruction.

NPR’s coverage of the report cited a U.S. study claiming teachers who use AI save an average of nearly six hours a week and about six weeks over a full school year. That number deserves scrutiny.

The "U.S. study" in question is a Walton Family Foundation/Gallup survey from June 2025 - predating the Brookings report, which cites it. The methodology was key -teachers who already used AI weekly were asked to estimate, to the nearest half-hour, how much time AI saved them on individual tasks. Those self-reported estimates were then summed. The result was 5.9 hours per week or, cumulatively, six weeks per year.

I don't buy it. The Brookings report itself hedges the claim in a section titled "Time savings only occur if teacher workloads remain steady" (Box 05 on p. 40 of the full report). More importantly, neither the Brookings report nor the Walton study can specify what is actually being done with the time saved. As Tom Daccord noted on his Substack:

Even if AI is saving teachers time, the benefits only accrue if that time goes back into individualized attention and deeper instruction. AI time savings might simply mean that teachers pack up and go home earlier. It might mean the time is absorbed into more administrative and ancillary tasks. The difference matters because if one of the major selling points for teachers using AI is to “save time,” then it is imperative that the time saved is used wisely. But I question the entire narrative that AI is saving teachers time in the first place.

The Walton/Gallup survey's own data complicates the six-week headline. Look at the frequency chart. Eighty percent of teachers have never used AI for one-on-one tutoring. Fifty-nine percent have never used it to make assessments. Seventy-three percent have never used it to supplement instruction. The "six weeks" figure comes from frequent users - a group that represents, at most, a third of the teaching population. Which means, for most teachers, AI time savings remain entirely theoretical. It’s also worth asking - if the time savings is so dramatic, why aren’t more teachers using it?1

The majority of tasks for which teachers are using AI are either tasks they were doing already or, in the case of something like modifying materials to meet student needs, ones which AI enables in a way that wasn’t feasible before.

This exposes a subtler problem with how time savings get measured. Consider a teacher who uses AI to build an elaborate differentiated worksheet in 20 minutes. She estimates that creating it on her own would have taken an hour and reports 40 minutes saved. But without AI, that worksheet never gets made. The actual time impact is plus 20 minutes, not minus 40. In that instance, the survey measures savings against a hypothetical task that would never have been undertaken. Multiply that logic across nine task categories and you get 5.9 hours of savings that potentially exists only on paper. The worksheet may well be superior. That's a quality argument, and a potentially legitimate one. But it is not a time-savings argument.

What the Workplace Data Actually Shows

Regardless of the inconsistency of self-reports, there is data available that undercuts the notion that AI time savings in the workplace translate into actual efficiency gains at all. Recent studies measuring behavior rather than perception tell a different story.

The ActivTrak 2026 State of the Workplace report, released in March 2026, analyzed 443 million hours of behavioral data across 1,111 organizations and more than 163,000 employees. Their conclusion was blunt: “The data is unambiguous: AI does not reduce workloads.” Among users comparing 180 days before and after AI adoption, time spent across every measured work category increased - ranging from 27% to 346%. No category decreased. Meanwhile, the average focused, uninterrupted work session fell to just over 13 minutes, down 9%, and continuing a three-year downward trend. The optimal band of AI usage turned out to be extraordinarily narrow - employees spending 7-10% of their work hours in AI tools showed the highest productivity of any group, yet only 3% of workers fell within that range.

A separate BCG study of nearly 1,500 full-time U.S. workers, published in the Harvard Business Review, also in March 2026, identified a phenomenon they called “AI brain fry.” The primary driver wasn’t AI use itself but the cognitive burden of overseeing its output. Workers with high AI oversight loads reported 14% more mental effort, 12% greater mental fatigue, and 19% more information overload. Those experiencing the condition made 39% more major errors and were significantly more likely to consider leaving their jobs.

These are not education-specific studies. But I’m not sure why we would assume the findings would be significantly different for teachers. Evaluating what AI produces is mentally demanding work, and the overall pattern across industries is consistent: AI adds new layers of work rather than reducing existing ones.

The Novice Teacher Problem

Imagine you are new to teaching. You’ve completed your training and you are organizing your course materials for the first time. What lesson should you start with? How should you set up your classroom? An AI chatbot will happily provide answers to all of these questions and some might be generically helpful. But which ones are right for your students, in your room, given your strengths? You have no way to know.

The ability to navigate the hundreds of decisions that go into teaching comes with experience - the kind that requires following your own instincts, getting real-time feedback from actual students, and adjusting the next day to adapt to your circumstances.

LLMs are great at generating ideas. Many are creative and potentially useful. An experienced teacher can scan them quickly and locate the one or two that might lead to a productive tweak. Newer teachers don’t have that luxury. They haven’t made the hundreds of mistakes that allow them to grab the one useful suggestion among dozens.

This is a judgment problem, and chatbots cannot impart judgment.

The “brain fry” finding is particularly relevant here. The primary driver of cognitive strain wasn’t using AI itself but overseeing the output. And that presupposes a major assumption – that teachers are actually carefully reviewing what AI produces. As models become more advanced and more verbose, the cumulative effect of reading and evaluating AI-generated text can easily take longer than starting from scratch.

Veteran Teachers Have A Different Problem

Even for experienced teachers, the time-savings pitch doesn’t hold up, though for different reasons.

Over the past year, I’ve written about several ways I use AI in my own practice - building a Claude Skill to audit lessons using the Understanding by Design framework, and creating interactive web modules for specific activities in my classes. Each of these was a significant investment that I believe paid off (or will pay off) in better student outcomes. But none of them saved time.

The lesson audit produced useful, actionable feedback. It identified misalignment between what I wanted students to learn and what my activities actually assessed. But evaluating every suggestion required 30 years of teaching experience. It was helpful but not “efficient” – at least not in the way the Brookings report envisions “time savings for teachers.” Performing an AI audit for every lesson during the school year is simply not realistic. That’s useful summer work.

The interactive modules are, I think, genuinely effective teaching tools. They are also time-consuming to build and require significant design thinking before AI can execute anything usable. My abiding mantra when creating or redesigning a lesson has been: what is the value add, and is it worth doing? AI hasn't changed that question.

Even for teachers who aren’t building custom tools, AI is still unlikely to save them time. Generating a quiz is easy. Proofreading questions, verifying outputs, and ensuring the product meets your classroom standards is not. Creating new worksheets from AI-generated ideas may be rewarding and great for your students. But the measure of “time savings” is always against the alternative – using the original lesson. If AI leads you to do something different and do it well, it’s a worthy goal, ideally one that leads to better teaching and learning. It’s not a shortcut.

Depth, Not Speed

The argument I am making is not that AI cannot help teachers teach better. Used effectively, AI is exceptionally good at helping teachers improve their own work. The newer models and capabilities offer possibilities for teachers with the curiosity and - most critically - the time to learn about and play with them.

All of the best use cases require significant up-front investment. Using AI effectively may also increasingly require access to the top-tier frontier models - an entirely different issue.2 These two points cannot be sugarcoated. AI opens up new possibilities precisely because it enables work that wasn’t possible before - not because it makes existing work faster.

The ActivTrak data confirmed what I’ve found in practice. AI doesn’t replace what we’re already doing. It amplifies it. But the claim that it saves time the way a dishwasher saves time or a calculator saves time is not what I've experienced.

Which brings me back to Drucker. AI makes it efficient to do things that shouldn’t be done at all - churning out generic lesson plans no teacher has examined, producing feedback no student will read, and generating worksheets that exist because they’re easy to create rather than because they serve a clear learning goal. If that’s the primary way in which teachers are using AI, that’s the Drucker problem in a nutshell. Efficiency without judgment is worse than inefficiency.

Time is the most important resource a teacher has. Be wary of anyone who tells you AI will give you more of it. What AI can do - for teachers willing to go deeper - is open up choices that didn't exist before. Those choices mean more decisions, and every one costs time and attention.

I’d love to hear from other teachers about their experience with AI as a time-saving tool. Am I wrong? What am I missing? Let me know your thoughts in the comments.

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

Schools should screen the new AI Doc - teachers, students, and parents should all watch this film. Preferably together. The reason is not because the documentary is especially illuminating on education. It isn’t. The current controversial and pressing questions about AI and schools are not the focus. This movie is not about AI and writing or cognitive offloading.

The reason the film should be viewed is to provide common vocabulary for us to have conversations about a technology predicted to be the most transformative in history. More importantly, it serves as an excellent primer for viewers unfamiliar with the existential debates surrounding AI acceleration and gives them exposure to both the apocalyptic and messianic positions those closest to the technology hold. Citizens, especially students, cannot challenge positions they don't know exist.

I recently saw The AI Doc: Or How I Became An Apocaloptimist with a colleague. The documentary arrives at a critical juncture. I recommend the documentary to everyone because it details the range of extreme positions taken by AI’s creators, risk assessors, and others in the field who are setting the framework for the debates.

What the Film Does

The basic conceit of the film is that co-director Daniel Roher1, an artist and Oscar-winning documentary filmmaker who serves as a stand-in for the audience, and his wife Caroline (also a filmmaker), are expecting their first child. Both are smart and likable - but Daniel's anxiety grows the more and more he learns about AI. He decides to investigate and make a documentary.

Some of the most recognizable names in tech agreed to sit and be interviewed in his makeshift studio, including Sam Altman, Demis Hassabis, and Dario Amodei, the CEO of OpenAI, CEO and Co-Founder of Google DeepMind, and CEO of Anthropic, respectively. The reaction shots of Daniel as he listens to one extraordinary claim after another are priceless - first the doomers predicting catastrophic outcomes due to unregulated AI development followed by a parade of AI optimists who tell Daniel this is absolutely the best time ever to be born. The doomers talk about AI leading to the collapse of humanity (or, as Eliezer Yudkowsky politely interjects, “abrupt extermination”) while the optimists describe a utopia where no one has to work, cancer is cured, and AI ushers in a world of abundance such as we’ve never seen.

For viewers who are hearing tech folks speak this way for the first time, Daniel is an everyman who articulates what the audience is thinking. Why are we doing this if these horrific outcomes are possible? How much and who am I supposed to believe about any of this? He falls into despair. Mid-film, there is an amusing and creative discussion with his wife via a cartoon flip-page notebook in which she begs him to become more hopeful as she carries their first child to term.

To watch the film is to accept the premise that AI is going to be a very, very, big deal.

But the hyperbolic venting of all these people is the point. In the back half of the movie, Roher comes to the quite reasonable conclusion that most of the issues associated with these fantastical prognostications are coming from the founders and developers themselves.

Altman comes off the strangest here in my opinion, partially due to the incredibly long pauses and cryptic answers he gives to some of the most direct questions. When asked if there are protocols in place for what happens if one of his models goes rogue and starts to do tremendous damage he says yes, but fails to elaborate on what they are. The fact that he is more than aware and clearly planning for the possibility is disturbing enough.

Why It Matters for Schools

Students are understandably tired of only hearing about AI in the context of cheating. Teachers are tired of harping on it. Education will be dealing with the fallout of AI for years, but the conversation is going to expand and only get more divisive and confusing as the technology continues to improve. I’ve argued previously that schools need to engage the larger issues about AI and The AI:Doc offers precisely that entry point. I can easily envision assemblies, advisory discussions, and guest speakers brought in to elaborate and answer questions about issues which, while they are going to affect everyone, will have a disproportionate impact on students enrolled right now. The purpose is not to accept the views of the film’s participants as gospel but to serve as a jumping-off point: for additional readings, other interviews, and opportunities to hear missing pieces and critical commentary.

Students will be rattled by the film. Public viewings will spur useful conversations that rise above the simple context of cheating and cognitive offloading that currently swamps the debate about AI and education. If this documentary is any indication, those issues are going to quickly become supplanted by much larger ones and I think everyone, but especially the younger generation, should gain access to the different sides of the debates.

These are the actual conversations a very small number of people are having when it comes to AI. What The AI Doc does is put forward a series of arguments and positions in front of the audience who may not be inclined to realize exactly how much and why Silicon Valley is focused on AI to the exclusion of almost everything else. There are statements made in the film that give the sense that the people who are closest to AI development lie awake at night contemplating the end of humanity. Others insist we're standing at the threshold of unprecedented human flourishing. Those are radically different perspectives.

There is an amusing sequence as Roher does a series of Google searches investigating how to survive off the grid if things go kaput, basically modeling how and why some people are turning into “preppers” for the world’s collapse.

The primary argument of the documentary is what might happen if we end up getting AI wrong, even if it approaches only a fraction of what’s being predicted, and how the general public should respond. The issues are significant enough that politicians, policy makers, educators, and anyone with an interest in the future for their children, need to be aware of what the people building these models think is at stake.

Yes, the Middle Position is Missing

Critics of the film will rightly point out that the missing middle conversation is left out. Yann LeCun and Gary Marcus, both well-known skeptics of imminent AGI, are not featured (Emily Bender gets a single quote).2 There is no exploration of “AI as Normal Technology.” Exponential AI scaling is taken as a given and the assumed timelines for AGI by most of the speakers are on the aggressive end of the spectrum. But I actually think that’s a benefit. It will be instructive to look back in 5 years to see whether this film was prescient or overblown.

One participant has invoked similarities to The Day After and I would also liken it to An Inconvenient Truth. The AI:Doc is a wake-up call. The people with the most direct knowledge of where AI is headed are genuinely alarmed, and quite a few are begging for government regulation - that includes the ones building it, like Amodei.

The sooner more people learn about these predictions, investigate these claims, and learn for themselves what’s likely to be true and what isn’t, the better. If screening the AI:Doc in schools has that effect, it can’t come soon enough.

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.

One of the best things about being a teacher is that work literally stops over break. Unlike other jobs, nobody needs someone to cover for you back at the office. I spent the better part of the past week away from a screen and it was exactly what I needed. It also showed me something I’ve known intuitively but hadn’t experienced directly - normal people are not following the AI news cycle with any kind of regularity. I am not normal.

So when I came back online this weekend, one story caught my eye. As of this posting, the story about Claude Mythos is a little more than 72 hours old. If you haven’t heard about it, join the club. While it will likely be lost in the deluge of the AI news cycle, it foreshadows an issue that will only become more relevant in the coming months - the staggering cost of running the most powerful AI models and who gets access to them.

How I Found the Story

My substack feed led me to Mythos through Carlo Iacono’s most recent post. It sparked my curiosity and gave me an opportunity to engage in the same kind of source analysis I model for students. How do you verify a story with so many moving parts and so little actual information to go on?

To find out, I turned to the SIFT framework, developed by digital literacy researcher Mike Caulfield.1 It stands for Stop, Investigate the source, Find better coverage, and Trace claims to their origin. Caulfield and Stanford researcher Sam Wineburg expanded on this approach in their book Verified.

What SIFT Looks Like in Practice

Here’s what I found when I traced the story back.

The original reporting came from Fortune’s Beatrice Nolan. Every hot take and every inflated claim on X flows downstream from those articles.2 What I would tell my students, however, is that even Fortune's reporting was based on documents independently discovered by two cybersecurity researchers - Roy Paz of LayerX Security and Alexandre Pauwels of the University of Cambridge - in a publicly searchable data store that Anthropic had failed to secure. Those draft blog posts and internal assets were the real primary source material. Anthropic locked down the data store after Fortune's inquiry, though by then cached copies had already begun circulating.3

Fortune’s reporting establishes the following: Anthropic has trained a new model currently called “Claude Mythos” (also referred to as Capybara), which is the next tier above its current flagship, Opus. The company confirmed it represents “a step change” in capabilities with “meaningful advances” in reasoning, coding, and cybersecurity.

The leaked draft is slightly more assertive than the public statement - it describes the model as “far ahead of any other AI model in cyber capabilities” and warns of “an upcoming wave of models that can exploit vulnerabilities in ways that far outpace the efforts of defenders.” Because of this risk, the planned rollout gives organizations early access to protect their systems before wider release.

The leak itself was caused by a content management system that made uploaded files public by default. Anthropic called it “human error” and said it was unrelated to any of its AI tools. It’s worth noting the irony that the company that has been warning about AI-powered cyberattacks lost control of its own documents through the kind of vulnerability its model is designed to detect.4

The article is based on internal draft “blog posts” which are essentially marketing copy written for a product launch. These are not internal memos or unfiltered emails. We're getting what Anthropic intended to say publicly, just before they were ready to say it. That's quite different from actually reading what they say privately and led to some commenters claiming the “leak” was intentional and simply a way to gin up hype. It wouldn’t be the first time a company engaged in such tactics, though Anthropic does not have that reputation.

That’s the gist of the verified story. Lots of people took this and ran with it.

Separating the Signal From the Noise

The Mythos story is a perfect example of a pattern all students and readers should know well by now. One verified story often generates a dozen unverified extrapolations. Within hours, new and unconfirmed details - ten trillion parameters, ten billion dollars to train - were circulating alongside claims that came directly from Fortune's reporting. One widely shared post mixed the two so seamlessly that a casual reader couldn't tell which claims were verified and which were invented.

That’s how it works - inferences, embellishments, and flat-out fabrications get shared more widely than the original reporting and errors get lost in the shuffle. Social media turbocharges the phenomenon and by the time most people encounter a story like this, the verified facts are indistinguishable from the noise.

The number one question students should ask when confronted with a claim that requires confirmation: can I trace it to a named source with direct access to the evidence? For the Mythos leak, the answer separates a surprisingly short list of confirmed facts from a much longer list of speculation.

One confirmed detail that matters most for educators was buried underneath the cybersecurity headlines. The model will be “very expensive for us to serve, and will be very expensive for our customers to use.”5

The Tiering Problem

Most students and teachers, if they are using AI at all, are using free tiers. Few people know which specific models power these free tiers, let alone how they differ from the most advanced versions. The default models on the free tiers of Claude, ChatGPT, and Gemini are all more powerful and multi-modal than what was available one year ago, and their performance is driving almost all of the AI and education conversation.

As the past few years have made clear, these base models are more than capable of disrupting the entire educational experience.

Yet these are not the models that make the boldest headlines. Each company now offers premium tiers - Claude Opus, ChatGPT Pro, Gemini Ultra - that are significantly more powerful and reserved for companies and individuals willing to pay top dollar for access.6 Meanwhile, the free tiers are tightening. Usage limits are getting stricter, resource-intensive features are being pulled back, and even paid subscribers are seeing throttled access during peak hours. The gap isn't just opening from the top - it's also closing from the bottom.7

What the Mythos story makes clear is that these frontier models are going to keep pulling away.

Opus, Anthropic’s current top-tier model, is already primarily an enterprise and power-user product. Mythos would add a tier above that and it’s explicitly described as too expensive for general release. The intense market competition among the three major AI developers - OpenAI, Anthropic, and Google - practically requires that the next model from each keep pace.

As AI models become even more powerful, they will be available first, and for an unknown period, only to organizations that can pay enterprise prices. Those organizations will not be public school districts. They'll be the European executives invited to preview unreleased Claude capabilities at an intimate retreat at an English country manor - a detail which was also in the leaked cache.

I wrote about this dynamic in February when the Moltbook story broke:

Mythos is Anthropic’s clear confirmation that’s where we are headed. For a while it seemed as if everyone was going to have similar access to the most advanced AI models available. By the end of 2026, I don’t think that’s going to be the case any longer.

The gap between what AI can actually do and what the average teacher or student has access to is about to get wider.

When we talk about AI in education, we tend to talk about “AI” as if it were one thing. As if the tool a student uses on the free tier of ChatGPT is the same as the one a top cybersecurity firm is using to audit its codebase. We talk interchangeably about the experience of a teacher using Gemini for the first time and the one a software developer has with Claude Cowork’s agentic process.

These are entirely different interactions with AI and it's one of the most overlooked dynamics in the education conversation. The tiering gap doesn't change what most teachers are dealing with right now - students can already do tremendous damage to their own learning with the free versions of these tools. We know this.

But when someone goes online and starts reading about what AI can and cannot do - as if each model is identical - it's clearer than ever why those conversations go sideways so often. Much of the commentary is based on premium models most users never experience, and the companies aren't exactly helping - OpenAI has quietly stripped model names from its interface, making the gap between tiers all but invisible.

The Same AI Experience? I Doubt It

It’s obviously too soon to tell whether Claude Mythos will be as consequential as Anthropic’s leaked promotional materials suggest. If the past is any indication, it will likely be another incremental improvement and garner attention mostly from within a small sphere of AI insiders. Anthropic has gotten more press recently, but it’s still a niche company. Most people you know have never heard of Mythos and probably never will. But if you’re someone who follows AI closely, it’s worth getting the facts straight. The newest and most powerful model priced at a level the average user cannot access means, if nothing else, that we are not all going to be having the same AI experience.

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and other institutions about pedagogical questions raised by AI. Take a look at my website and reach out - I’d love to hear what you’re working on.