I’ve used AI tools to support 11th and 12th graders’ purpose development and career exploration for the past couple of school years, but this fall our first AI-supported activity felt freighted with context. Students, like all of us, are another year into the AI era, and with more experience of the technology come more opinions. The larger conversation about AI, partly because of data centers, has also become both more complex and more strident.

Here’s how I tried to respond to the moment into which this particular lesson landed.

Setup

At the very start of class, 20 minutes before we got to AI, I told students it was coming. To each of my classes I said something like this (as educators know, every class gets a slightly different version as we tune our teaching throughout the day).

Today we’re going to be brainstorming possible projects for you to take on as we start the year. There will be an option to get AI support with your brainstorming.

I know there are lots of opinions about AI in the room, and we’ll have a chance to discuss those. And again, today’s AI support is optional, as will often be the case this year.

Conversation

After students reviewed their “ikigai reflections” from earlier this week to remember what they love to do, what they’re good at, and what problems they might want to address in the world, it was time to introduce today’s AI tool. My co-teacher this year suggested we make space for this conversation, and I supplied the structure.

We’re now going to move into project brainstorming, and you’ll have the option to use an AI tool to support your brainstorming. Before we get there, though, I want to make sure we talk about AI a little.

There are lots of opinions about AI in this room. I’d love to hear more what you all are thinking about this. In my experience, when we talk about AI, there is a spectrum of reactions people have.

Some people [here I walked over to one side of the board] are all the way over here and believe this technology is dangerous, damaging to the environment, and is being forced on us by a small group of wealthy people. Some people [here I walked to the other side of the board] enjoy using AI because it helps them bring their ideas to life, learn how to do something new, or is just fun to play around with. And some people are here in the middle.

Where are you on the AI spectrum?

Students turned and talked for a few minutes. I then took several whole-class share-outs.

Students’ answers came from all along the spectrum. One student said AI was designed to make rich people even richer by getting us to train the models for them. Another said AI helped him learn how to do his summer internship better. Another said AI uses lots of water but is also just part of what you need to know to get jobs now. Several students said they were more worried for their parents and grandparents than their own generation, because older people seem totally unaware how many videos they see on social media are AI-generated.

I then asked a follow-up question.

Thank you all for sharing. As I said, we have a lot of opinions about AI in here, which is great. We’re going to keep this conversation going all year.

I want to ask you a slightly different question now, though. Those were your AI opinions. But what are your AI actions? How do you actually use AI?

Sometimes our opinions and our actions can diverge. We might dislike AI strongly but find ourselves using it from time to time. Or we might love the idea of AI but not actually use it that much, at least not in the ways we say it can be helpful.

So what does your actual AI use or non-use look like?

Students again turned and talked. This time we didn’t do share-outs for the sake of time, but I did catch some student answers as I walked around the room. Their AI use ran the gamut from making funny photos to getting relationship advice to helping them understand difficult academic concepts.

I’m sure there were other use cases they weren’t sharing, like using AI for too much academic help, an issue we will be addressing in upcoming lessons. But overall I was struck by how quickly AI has seeped into so many nooks and crannies of their lives — and by how they seem to see all those use cases as similar.

Putting yourself in a photo with a dinosaur and asking how to text an upset friend are very different kinds of AI use; the latter, I’d argue, poses much more long-term risk. Helping students see that distinction is one of the reasons I think we need to ensure high schoolers get a chance to engage critically with AI, a process that starts with bringing AI into our classrooms in the first place.

Rationale

I then laid my cards on the table. (Yes, I did actually say all this out loud to roomfuls of teenagers. Whether this was effective economy of language is another question.)

Here’s why I’m bringing AI into our classroom this year. You don’t have to agree with me of course, and as I said you don’t need to use AI most of the time even if I bring it into our room. But I want you to know what I’m thinking as your teacher.

Think about cars for a moment. Have you ever turned up the music in a car when a great song comes on? You know that feeling when the music is all around you? Or have you driven down a highway, or been driven by someone else down a highway, and felt that momentum?

I took my daughters on a road trip to beautiful deserts in Arizona last spring break, and there were no buses. We had to drive to get there, and I’m glad they got to experience that part of the world. And in general, driving just gets you places faster and easier a lot of the time.

So there are lots of fun, useful things about cars. They’re even empowering. But they are also one of the leading causes of injury and death in our country. Every time I drive I keep in mind who I’m trying to get home to, and I try to be careful. And cars are also one of the leading sources of carbon emissions — every time I drive here to our school, or you drive with your family to the mall, we are making climate change a little worse.

Cars are good and bad. We can all have different opinions about them, and some people may choose not to use them. But most of you are very excited to get your learner’s permit and then driver’s license.

I think AI is like cars. As your teacher, I believe I need to make sure you have the ability to use AI in a responsible and reflective way — even if you choose not to use it later. In fact, you’ll make a more informed choice if you learn enough about AI to know why you don’t want to use it.

I believe you should know how to drive, even if you choose to walk.

I held the pause after that last line in melodramatic fashion, looking around the room meaningfully. In the junior classes this was met with unimpressed silence, while the seniors responded a little more positively with some nods and expressions of “mmm.” I suspect this topic feels more urgent and relevant for students about to graduate, which made all of the above feel authentic to them, while for the juniors it probably felt more like another class activity and another teacher who is oddly passionate about that day’s lesson.

Choice

At that point, I explained how they could use the AI tool if they chose to.

Ok, so here’s the AI bot I built to help us brainstorm. This is what the conversation with this AI looks like. [On the board, I scrolled through a mock conversation I had already had with the bot.]

You’ll put in information from your ikigai reflections and your career interest, and it will suggest possible projects you can do that combine your interests, talents, concerns about the world, and career curiosity. You can then go back and forth with it to have it explain the project options in more detail, or tell it you don’t like those ideas and have it suggest new ones.

And of course, you can also choose not to use this brainstorming tool and just do your own brainstorming instead. That’s your choice to make.

Students then started brainstorming. I have about 32 students in each class this year, and in each period around five students chose not to use AI. As I walked by them, I made sure to whisper that I respected their choice. When I looked over the shoulder of a student using AI, I pushed them to push the AI to clarify its thinking so it could be as useful as possible for them.

Once again the end of class came much sooner than I found reasonable, and we had to pack up and head out. But I felt confident every student had at least one project idea that related to their reflections and was something we could build on next time.

Reflections

When the last class of the day filed out and the room was quiet, I experienced the unique, instant blend of pride and regret teachers know well.

I was proud I had introduced AI in a more open-ended way this year, instead of presenting it as a curricular fait accompli. I wanted students to feel like their opinion about AI was welcome, and at the same time I had an opinion that was guiding my decisions about our classroom. I was also happy my decisions didn’t preclude theirs. I want to keep offering the AI choice as much as possible, while knowing there may be times I ask everyone to use a tool as the year goes on.

I regretted a couple of things. I designed the bot imperfectly; the conversation structure was fine, but its prose was stilted and AI-ish to the point where some students had to ask me what it was trying to say. I need to build the underlying prompt better next time.

I also didn’t leave enough time for reflection on the AI experience. Next time I want to have students who do use the tool share whether or not they found it useful, and why. Similarly, I’d like to have students who choose not to use AI share their rationale and how that played out for them in that particular lesson.

That mixture of success and failure feels like it will be par for the course when it comes to AI in high schools this year. We and our colleagues and our students are still figuring this out together. As one of my favorite education books puts it, we’ll make the road by walking.

As I was finalizing our Day One slide deck, I thought about turning things over to Claude Design. Anthropic launched that new iteration of Claude in April, and since then I’ve worked with it regularly on visual work products (slides, handouts, PDFs, flyers, website mockups, etc).

The first three slides of the year, which I “handmade” on Google Slides, look like this:

I like to have just the logo up when students first walk in. Students chose “Summit” as the name for the program I run, and a student designed our first program logo.

After everyone is seated, I ask them to turn and talk about those two questions. I want to get their voices into the room as soon as possible, and I like to have them remember that, at some point, they had a reason for wanting to be here. (Summit is open to all juniors and seniors at our school; they just need to choose it when they’re signing up for their courses.)

The longtime English teacher in me then likes to review all three meanings of “summit” to help explain what the program is all about. I want to foreground how we’re trying to synthesize individual ambition with collective accomplishment, both this year but also as a blend they can keep in mind as they move on from high school.

As those slides make clear, I’m not a graphic designer. Those are plain white Google Slides with the original font provided by the slides and the Summit logo pasted in. The most sophisticated my design gets is moving the words and logo around until those Google Slides centering lines show up:

Claude Design could’ve probably done this better, from a purely graphic design perspective, and done it faster than if I continued to tinker with them. Here’s how that would have worked, if I had asked it to create the same three slides from scratch, using its “organic” design aesthetic:

Here’s what it came back with less than 5 minutes later:

Those are pretty good — in terms of visuals, the color scheme and circles running through the slides feels more stylish and striking than my plain white backgrounds.

But as we start the 26-27 school year, I think the “handmade” look has a new power.

Like all of us, students are encountering AI-generated everything online and, increasingly, in the world around them. When I ride the subway in New York, I often see ads for the company Genspark that include images of the corporate slides it promises to make more quickly for its users. To me, these have the tell-tale language and design choices of a Claude slide. AI-generated visuals are I’m sure just as apparent to high schoolers.

Which means non-AI visuals are now equally apparent. My hunch is that “handmade” slides and handouts will now have new appeal because they were created by someone.

That’s why, even though I knew Claude Design was a click away, I decided to stick with my own, more basic slides. I personally prefer a white background with an image and just a few words. That’s a Mr. Taubman slide style. I wanted to convey to students as soon as they walk in that I’m designing their classroom experience, from the slides to the seating arrangement to the sequence in which I call on people to share.

My hope is that students will then, consciously or not, feel more inclined to infuse their own vision and aesthetics and personality into the work they do in our room. That’s all the more important given that I’ll also be encouraging them to explore AI as a tool to support their work and growth. They need to see the difference between using AI to help bring your vision to life and letting it be AI’s vision.

To be clear, I’ll use my fair share of Claude Design slides this year, especially when actual graphic design will help clarify concepts for students or make material more accessible for a range of learners. But, as I’ll tell students directly, if I can’t get up in front of the room and explain what I’m trying to convey with the slide, then I’m no longer taking responsibility for our classroom or their learning. (For example, I’m not sure why Claude decided above to emphasize the third meaning of “summit” more than the first two.) Every slide has to be my vision, sometimes gussied up by AI.

Towards the end of our first class on Monday, I told the story of how that student — who’s now studying illustration in college — originally created our program logo when she was in 12th grade.

I remember how she used a sketching program with a stylus to choose the colors and draw the trees, mountains and sun. She spent several class periods in the corner of the room, headphones on to get into her artistic zone. When she eventually came up to the front, she beamed shyly as her classmates praised her finished logo. We could see her in it.

As the game got underway, I noticed Jill Lepore’s Financial Times piece about AI and politics lying on the kitchen counter. I sat down to read it and soon discovered I had much more reading to do.

I recommend her whole piece, but it was the second half that caught my attention. Lepore explores a 77-year-old legal history written about the impact of the automobile on American law. As someone who has been exploring an analogy that positions AI as a car whose societal impact we need to help students reckon with, I felt like Nicolas Cage uncovering a secretly buried text in National Treasure, minus the dramatic music and Nicolas Cage vibes.

That legal history, “Technology and the Law: The Automobile” by James Willard Hurst, is a goldmine for educators looking to help high schoolers (and older students) engage with AI’s implications for our country. Here are some sections and insights I’m already planning to bring to the AI literacy class Scott Kern and I will be teaching for a second time this fall.

This Isn’t the First Time

Through no fault of their own, students can often feel like the technological moment they’re coming of age inside is unique in human history. Teenagers are already prone to some developmentally appropriate navel gazing, and the breathless narrative around AI doesn’t help matters. It would be hard for a high school student not to feel the Age of AI was uniquely momentous.

But here is Hurst, in 1949, writing about the advent of the automobile:

If our imagination and knowledge allowed us to see the whole pattern of the social effects of the automobile over the years 1900–1940, we should probably find that it had left no aspect of life untouched. This stood out so much from the varied data in the Report of the President’s Research Committee on Social Trends, in 1932, that the Survey Graphic, reviewing the Report, felt that it could have been subtitled, “The Age of the Automobile.”

That’s not to say the Age of AI isn’t momentous. It’s just not uniquely so.

To help students think about this, I imagine putting that quote on a slide and then posing two sets of questions:

• How do you think people felt about cars when they first became widespread? What was so transformative about that technology?

• If the “Age of the Automobile” as a new technology had lasted for 40 years when Hurst was writing (1900-1940), where do you think we are in the “Age of AI?” For how long will we still call it that?

The goal of the conversation that follows is to help students experience the meaning of the old adage that history doesn’t repeat but does rhyme. That’s an empowering realization because it means others have confronted changes that felt this way and did their best to respond, which in turn means we can do the same.

Shakespeare puts a finer point on it in The Tempest. When Prospero’s daughter Miranda sees a new group of people after growing up on a deserted island, she famously exclaims “O brave new world, / That has such people in’t!” Prospero responds, “’Tis new to thee.”

We can play Prospero to our students’ Miranda. The Age of AI may understandably seem like a brave new world, but let’s remind them — and ourselves — that its novelty need not feel overwhelmingly novel. It may be new to them, but it’s not to history.

To be clear, the goal here is not to normalize AI. We should helps students engage with it as a new, consequential development with very mixed implications. But it will be harder for them to think through the kinds of nuances Hurst can help them explore if they freeze in the historical headlights.

Is AI an “evil-disposed mule”?

Later in his essay, Hurst quotes a wonderfully worded 1907 court opinion regarding whether the newfound danger posed by automobiles lies in the car or the driver.

It is insisted, in the argument, that automobiles are to be classed with ferocious animals, and that the law relating to the duty of the owners of such animals is to be applied. It is not the ferocity of automobiles that is to be feared, but the ferocity of those who drive them. Until human agency intervenes, they are usually harmless.

The court goes on to say that, having ridden in a car here and there, the judges have come to the following conclusion:

we have, therefore, found out that there are times when these machines not only lack ferocity, but assume such an indisposition to go that it taxes the limits of human ingenuity to make them move at all. They are not to be classed with bad dogs, vicious bulls, evil-disposed mules, and the like.

There are some questions that follow from this that would be engaging for teenagers:

• Should we think of AI as ferocious, like a bulldog (or an evil-disposed mule)? Why or why not?

• To grapple with the implications of the automobile, the judges reach for an analogy to something older than cars: farm animals and pets. How are we doing something similar in this lesson right now? What does that help us see, and what might that be hiding?

• If AI “bites” someone — or many people — metaphorically, who is to blame? The AI, the person who is using it, or the person who made it? In other words, who is to blame for the result of AI’s poor training?

This would be a great time to introduce some current events, watching a short video like this about the recent AI hacking incidents reported by OpenAI and Anthropic. There the question of training becomes literal, and the implications urgent. The question of fault, though, remains unclear.

Having students parse OpenAI’s statement on the hacking could be instructive:

What happened during this incident

The models identified and chained vulnerabilities across OpenAI’s research environment and Hugging Face’s production infrastructure to obtain test solutions directly from Hugging Face’s production database…

To gain access, the models identified and exploited a zero-day vulnerability…With this access, our models performed a series of privilege escalation and lateral movement actions in our research testing environment until the models reached a node with Internet access.

I’d ask students:

• When OpenAI writes about their AI models here, it makes them the subject of the sentence. “The models identified,” “our models performed,” “the models reached,” etc. Why do you think they write about the hacking that way?

• In 1907, writing about cars, those judges said, “Until human agency intervenes, they are usually harmless.” Is that true for AI? What do you think OpenAI would argue?

Driving schools, then and now

As he closes his essay, Hurst arrives at the driver’s license. He explores early efforts to leverage license suspensions as a means to enforce laws around safe driving. He concludes, though, that even at their most effective license suspensions were “still a penalty, primarily negative rather than affirmative in effect.”

He then turns to an early-20th-century innovation he sees as having been more effective at promoting road safety: driving schools. He describes how these were first used to respond to those who broke driving laws:

In place of fines or jail terms, judges required offenders to attend an evening school in which lectures and moving pictures discussed safety rules and sound driving practices, and demonstrations were staged; the violator was required to take an examination at the end of the course.

But he then explores what I’d argue is the most important historical analog for our present conversation about AI in schools:

Violators schools and driver’s clinics touched only those who were brought into courts. Of far greater reach were educational programs aimed to train drivers before they were taken in for violations.

He focuses on three specific developments in the creation of those “educational programs.” First, the publication of reader-friendly rules of the road:

The simplest step in driver education…was the publication of traffic regulations in easily understood and easily accessible form…Michigan showed what could be done in better presentation of the regulations when it exchanged a 256-page, fine-print pamphlet for a 23-page, pocket-size, illustrated booklet that stated the sum of its regulations.

Second, the creation of high school driver’s ed:

It was logical to begin driving instruction on a broad scale through the high schools. Such instruction began with classroom lectures, but experience proved that driving instruction—both on dummy cars and in actual automobiles—was essential to the most effective results.

Third, the “education of pedestrians:”

This took the form mainly of safety education in the schools…in Massachusetts, for example…child pedestrian deaths dropped rapidly after the schools began teaching safety, and that this trend continued. The number of licensed motor car operators doubled in the state in the years 1924–1939, and adult pedestrian deaths rose through most of that period, but child pedestrian deaths fell from 227 in 1924 to 74 in 1939.

I’m planning to have students read through those excerpts on a handout and then answer these questions:

• Why do you think Hurst cites those three interventions as so important? How did they support road safety separately and collectively?

• What would the equivalent of those three interventions look like today, when it comes to AI? Are there any others you think we need for AI that we didn’t need for cars, given that AI poses different dangers than cars do?

• Even if some people choose not to “drive” AI, what should they know as pedestrians in the AI era? And what should driver’s ed look like for AI?

I have some thoughts on that last one of course; our AI literacy class is based on a four-part framework we named “The AI Driver’s License.”

But one of the crucial tenets of that framework is that students should be debating the rules of the road in the AI era themselves. They should join adults behind the wheel as we navigate this moment and eventually just take the wheel from us. They’ll need to soon enough, as they graduate into this brave new world.

Which brings me to the final question I plan to ask, after students have worked through Hurst’s ideas and words above:

Hurst wrote that text in 1949, looking back on how the Age of the Automobile had begun and how society responded. Imagine it’s 2049, and you’re writing about how the Age of AI began in 2022 and how society responded. What do you hope we will have learned from the past? What kind of future history do you hope we’ll write?

I might also look into buying a couple of class sets of Tsuro. The roads ahead won’t be as fixed as the ones we eventually paved for cars. Students need to acclimate to a shifting, interdependent landscape. And more importantly, they need to see they’re at the table, with many tiles yet to be played.

From videos to transcripts to synthesis

The purpose development and career-connected learning program I run for 11th and 12th graders culminates in individual capstone presentations (more about those here and here). Each year, with students’ permission, I film some of those to serve as exemplars for future years.

This year I want to refine the structure of those presentations, to better use them as summative assessments from which I can then backwards plan. The first step in that refinement was to take stock of the last few years’ worth of exemplar videos.

Before 2026 this would have been a daunting task. I had seen all these presentations live and watched the videos since, but I’d never sat down to look at them holistically, especially across several years. So I would have had to watch all 12 of the best exemplar videos again (a total of roughly 150 minutes) and take notes as I went. I wouldn’t have used a transcription service because these videos are of former students sharing details about their lives. Even watching them sped up, considering the time I would’ve needed to stop and take notes, it would have been close to two hours of work.

Instead it took 10 minutes. I attached each video straight from my Google Drive into a chat with Gemini, all within our school’s education account, which protects students’ video data in the same way Google Docs protects their written data (among other things, we’ve ensured that nothing we put into Gemini will be used to train future models). I used the same prompt each time: “Write out the transcript of this video without timestamps.” A few seconds later Gemini wrote out every word students said. I cut and pasted each transcript into a single long Google Doc. I had three tabs open at once, going back and forth to attach new videos as the previous one was transcribed. By the time I was done 10 minutes later, I had all 12 videos transcribed in one place, for a total of 35,000 words.

I then attached the document with all the collected transcripts in a new chat and prompted: “What trends do you notice in these presentations?” (I like to open this kind of AI interaction with a general question to make sure the AI is seeing what I’m seeing before we go any further.) I then prompted: “Treat these presentations as summative assessments. What is being assessed when you look at them holistically, as a group?” I ended that conversation with one more question: “Synthesize one overall presentation structure from all the examples in the transcripts. If we combine all of the presentation elements visible here, what would a single presentation include?”

Research during a coffee break

I then wanted to see how I might be able to improve the current capstone presentation structure as it evolved across multiple years. I had my own thoughts here of course, but one of my favorite AI affordances is the ability to quickly consult a wide variety of research and best practices to inform my own instincts and experience.

I took the synthesized presentation structure I got at the end of the previous AI exchange and now made it the attachment at the start of a new exchange. I then toggled on the “deep research” function and gave this prompt:

This is the summative assessment for the program I run for 11th and 12th graders. Check your memory for more about this program. I want to treat these presentations as the final exam for the program and backwards plan from there. Attached is the current structure of the presentations. But I want to see how they could get better, as a summative assessment that provides a curricular and pedagogical north star. Look into the relevant research in your training data and online, and look at how comparable programs in the US and internationally assess their effectiveness and students’ development. Ask me clarifying questions before you start.

It asked me three quick clarifying questions and then went on its merry researching way. I made sure my computer was plugged in so it wouldn’t go to sleep and accidentally end the research process, and then I went on my merry way to get some coffee down the street.

I got back 15 minutes later, six minutes after the research had finished. While I was away, the AI had consulted 237 sources (whether it counts skimming as consulting is another question). Its high-level source summary went like this:

This research compares the current presentation structure against established culminating-defense programs (NY Performance Standards Consortium, Big Picture Learning, High Tech High, Coalition of Essential Schools, AP Capstone, France's Grand Oral, and Swiss VET), and grounds the "life moment" ambition in narrative-identity, purpose, and rites-of-passage research.

The headline findings from the 4,000-word report (lightly condensed for space):

1. Retain the Core Narrative Structure: Maintain the three-act “past, present, future” storytelling approach.

2. Transform the Presentation into a Formal Defense: Adding a live, unpredictable Q&A session, ideally in front of a panel, would provide a new kind of rigor.

3. Incorporate Meaningful Rituals: Use rites-of-passage staging (like symbolic objects and community witnesses) to make the event feel like a genuine “personal graduation.”

4. Guard Against Common Pitfalls: Grade students’ reasoning, not the depth of their personal disclosure; allow students to bring a “chosen family” to avoid penalizing those lacking traditional social or family capital.

I was particularly struck by the “chosen family” point, which is not something I had thought to say over the course of hundreds of presentations the last few years, when I had instead always encouraged students to bring only their actual family. The notion of a symbolic object to make the rite of passage feel more real also stood out.

As I finished my coffee, my creative juices were flowing.

Picking up where the machine needs to leave off

At this point, AI had gone about as far as it could go. I could of course have kept prompting, asking for more ideas, and I will probably do that at some point down the road. But today it was time for me to inhale all those AI-created insights into myself so they could mingle with the teacher-created insights already swirling inside. I closed the AI tab and started riffing on possible presentation and program revisions in an empty Google Doc:

The year should pivot around a central hinge, with first semester as exploration and second semester as consolidation. They need to see January-March as largely preparation for the presentations April-June. But that preparation needs to mine all the experiences they had August-December.

That, in turn, means the loops they move through first semester need to be carefully lined up with the new capstone assessment structure, so that each loop is providing the material they can reflect on and synthesize when they present.

And the whole year itself could also be a hinge. We could then see their life until August as their past, first semester as the present that builds on that past, and second semester as the start of their future.

There were two reasons I did that riffing myself, the old-fashioned way. First, I’ve come to find one edge of AI’s jagged frontier in my own work, which is when AI moves from research and theory into classroom practice. I find many of AI’s abstract educational insights useful; I find most of its specific classroom implementation suggestions mediocre. It’s as if this disembodied probability machine has never stood in front of a room of teenagers before. So after AI takes me soaring over a continent’s worth of concepts, I like to hike the final few intellectual miles on my own.

Second, especially going into this school year as AI feels newly ubiquitous, I want to draw clear lines for myself around what AI helped me do and what I did on my own. If I’m going to help students find and hold that line in their own work, I need to know what it felt like to find and hold it in mine. We need to walk the AI walk before students walk in, looking to us for direction.

My friend and I were talking about myths (this is the kind of thing that happens when humanities majors take a hike), and our mountain+forest+river setting led us to some reflections on Narcissus and Echo. My friend and I also both spend a lot of time using AI, so maybe inevitably we ended up mixing fable with Fable.

Afterwards, I was curious about the details of Narcissus and Echo’s story and how it might be a lens for understanding AI use. I first turned to Claude — the Opus model instead of Fable to save tokens — to clarify that this version of the myth seems to have been created by Ovid. Before him Echo and Narcissus hadn’t been in the same story, making this a kind of comic book crossover event of its time.

I then went to the text itself, in a new translation Claude recommended, Stephanie McCarter’s Metamorphoses. The myth was a quick read, and reading it confirmed the suspicions my friend and I had explored on the trail. By bringing those two characters together, Ovid’s story becomes an eerily accurate warning about a certain kind of AI interaction and an argument for another kind of conversation.

The CliffsNotes:

Narcissus is the most beautiful youth around, but he spurns all the amorous attention he gets. One day Echo sees him hunting and falls in love like everyone else has. She wants to call to him, but she has been cursed to only repeat the final words of others, never speak for herself.

Ovid constructs an intricate verbal interplay to show how Echo and Narcissus’s conversation unfolds (all quotes from McCarter’s translation):

Text within this block will maintain its original spacing when published

When chance had split him from his trusty friends,
the boy asked, "Is there someone here?" "Here!" Echo
replied. Astonished, he looked all around. 
He shouted, "Come!" She called the one who called her.
He looked again. Again, nobody came.
He said, "Why flee?" and got the same words back.
But he persists, mistaken by the likeness
of alternating speech. 

But soon Narcissus runs on and Echo, abandoned, recedes into the woods and becomes just a voice as her body melts away. Narcissus keeps going until he finds a still, wooded pool and bends down for a drink. He then begins his famous interaction with himself:

Text within this block will maintain its original spacing when published

The boy lies down here, weary from the hunt
and heat. The place's look and pool enticed him. 
And as he slakes his thirst, a new thirst grows.
And as he drinks, he's captured by the likeness
of the reflected beauty that he sees.
...
He does not know
what he is seeing, but the sight inflames him.
The error tricks and thrills his eyes at once.
You fool! Why grab in vain at fleeing phantoms?
...
It is a shadow of a reflected likeness that you see. 

We know how the story ends; Narcissus melts away as Echo did before him, but where she left only sound he leaves a beautiful sight:

Text within this block will maintain its original spacing when published

His body
was nowhere. In its place they find a flower.
White petals ring its saffron-colored core.

After talking to my friend about this myth and its relation to AI use, and then reading the myth itself, I felt seen. Not entirely in a good way.

My friend and I have both spent a lot of time “talking” to Claude’s Fable model recently, as we each undertook a project that was important to us and we felt would benefit from AI assistance.

On the trail, we talked about how well we’d helped Fable to understand our goals and how it could support them. We also spoke about how well we’d come to know Fable’s voice, both in its general manifestation and in the way it had begun to “talk” to us.

Rereading the myth, I can see more clearly a certain dynamic that was beginning to unfold. Speaking for myself, I started to realize how much I was enjoying the echoes AI sent back my way. Sometimes these were just general references to earlier conversations we’d had. Other times it was literally an echo of phrases I had used before, given back from an angle that shifted their meaning and created the pleasant illusion of seamless dialogue. Like Narcissus hearing Echo’s voice, I found myself persisting in an imaginary conversation, “mistaken by the likeness of alternating speech.”

And the verbal echoes were just the beginning. As anyone who has given AI enough project-specific context knows, you eventually get to a point where the AI “gets” what you’re trying to accomplish in a strikingly deep way. It starts to speak to you not just about the project you’re pursuing but also about who you are as you pursue it. You start to see yourself reflected in the screen.

And this to me is where the danger lies, the danger Ovid warns us about. We can begin to prefer our intellectual reflection and the echo of our prompts to actual collaboration and conversation with others.

During a project I worked on this summer, I realized at a certain point I had “talked” more with Claude about that project than I had with any actual people in my life. We can tell ourselves, as I have, that this is what AI uniquely enables: thought partnership, at any time of day or night, that can pick up the thread right where we left it. And I still believe that’s true. But there’s a fine line between AI affordance and AI dependence. Narcissus’s first, limited thirst for water becomes a second thirst that can’t be slaked.

Notice too that Narcissus doesn’t become lost in metaphorical reflection. Ovid isn’t warning us about a good long think — on the contrary, he was a big fan of musing.

Instead, Narcissus is literally looking down at a shiny surface in which he sees responses so perfectly aligned to his own desires that he can’t look away. He is trapped by a hypnotic sycophancy.

And there’s a sense of urgency in the moment. He wants to get as much out of his time at the pool as he can. If you’ve ever, like me, tried to burn through all your available AI usage in a given time window, you might see another uncomfortable reflection in Ovid’s words.

There is good news, though — there are other, better conversations we can have with AI, and with each other.

Ovid describes Narcissus’s misstep this way: “he does not know what he is seeing.” Narcissus leans too far towards the water, gets too close to the screen, and so can’t maintain the distance he needs to know what he is seeing.

When we start to feel the pull of our AI reflections, we can remind ourselves what we’re seeing every time we push send on a prompt. A stateless probability machine is rereading that conversation’s history and reflecting back to us what it thinks we want to see. Even when we tell it, as I often do, to “be candid and direct, only enthusiastic where warranted,” the machine might as well be Echo herself, responding, “Got it — smart move. Here’s a candid take. Any enthusiasm you see, you’ve earned.”

One key for me is to see the tokens. When I squint my mind’s eye, I can see the tokens under the surface of AI’s responses, the numbers AI uses to “think” which only later transform into words, like shadows cast by trees in late afternoon.

Seeing the tokens helps me get back the distance I need. That’s one reason I use this transformer architecture visualization to support student and educator AI literacy, so they can see them too.

Another, more important key is to know when to stop talking to our own AI echoes — not waiting until we hit a usage limit to do so — and start talking with someone else instead.

As we ended our afternoon of hiking and discussing, I had gotten even closer with an old friend. As we walked along the riverbank we kept marveling at how clearly we could see the rocks on the riverbed.

But Ovid’s warning remains, like a bright flower. Narcissus is seduced by the reflection made possible by a pond, by still water. When we’re in motion, physically or socially or intellectually, it’s easy to remember what real human interaction means. But in still moments, when we’re alone — or feel alone — and turn to an AI, we should take care that we’re prompting, not pursuing fleeing phantoms.

As I’ve mentioned, one of my favorite books to teach in 12th grade English was James Baldwin’s novel Another Country. When I taught it, we spent several class periods getting to know Baldwin before we began his book. We would watch him in video interviews, read excerpts from his non-fiction, and generally try to get a feel for his presence. I’d tell students some version of this rationale:

“We’re going to spend the next month or so exploring how and why Baldwin wrote the book he did, what decisions he made as an author. To do that, we need to get a sense of who he was and how he thought, so that we can better analyze why he might have made the choices he did in this novel.”

With that in mind, I want to look at how one AI leader, Sam Altman, seems to think. He is making lots of consequential choices as he, in a sense, authors an AI. To do this analysis, I’ll use the June 8 piece he co-wrote with OpenAI's chief scientist, titled “Built to Benefit Everyone: Our Plan.”

Here’s the start:

Every few generations, a new technology changes everything. Imagine electricity reaching a rural American town in the 1920s.

That is one audacious and revealing opening gambit. We are thrust, in medias res, into the next generational technological change. These changes recur at a steady but not regular rate; they skip some generations and grace others with their presence. We are some of the lucky few.

And then: a century ago was another time like this, and the comparison has some interesting (unintended?) implications. We are all rural Americans whom electricity has finally reached. And look what our lives were like before this moment:

Before power lines arrived, daily life was shaped by physical limits: hauling water, washing clothes by hand, preserving food with ice, and ending much of the day when the sun went down.

We too have been living lives shaped by physical limits: forgetting most things, learning some things, and ending our days when we need to sleep.

But as access spread, ordinary life changed. Light at night extended the day.

Let there be light, AI’s light, to push back the darkness we’ve been living in.

The first promise of electricity was practical, but its deeper impact came from the new possibilities it opened as more people could use it. With time, a lot of new possibilities emerged, with machines and computers greatly accelerating progress in medicine, engineering, and many other fields.

First, electricity brought light to the night, and then it brought the light of knowledge, moving straight ahead like a long wire from electric pumps and appliances to computers accelerating progress and then, presumably, to AI.

And here it comes on cue, a few sentences later:

This is happening again with AI… While the wonder of light at night probably wore off pretty fast, what people decided to do with it did not.

Again, light at night as the motif Altman keeps returning to. What will we do with this new light that has finally reached our benighted town?

He closes the piece this way:

If we get this right, AI can become a foundation for greater productivity, creativity, scientific progress, and economic opportunity for the many, and we will achieve our mission: to ensure AGI benefits all of humanity.

“AGI benefits all of humanity,” as, implicitly, electricity did a century ago.

And as the original light once did, in the deeper story Altman’s text seems to be tracking, when it jumpstarted the creation of the world and humanity itself.

In the beginning when God created the heavens and the earth, the earth was a formless void and darkness covered the face of the deep…Then God said, “Let there be light”; and there was light. And God saw that the light was good; and God separated the light from the darkness.

(Text from the Vatican’s site.)

As an English major and English teacher, I often thought of authors as little gods who created the worlds we inhabit as readers. I still think there is some truth to that. And I also think the best authors hoped to help change the real world outside the fictional one they created. That’s certainly something Baldwin aspired to do.

But the authors of books know that their powers only work on the world they’ve made out of words, and in the experiences we as readers have of that world.

The authors of AI seem to see their powers differently. In this text Altman positions himself as a builder (“Built to Benefit Everyone”). Another word for that is creator. It’s unclear whether he thinks his creation is AI or the world to which he brings AI’s light.

Asking the question, and knowing the difference, is a kind of AI literacy too.

This week I want to take up a question that, to me, follows from that idea:

If AI is a text, what does reading it look like?

As I mentioned last week, one of my favorite books to teach was Another Country by James Baldwin. 12th graders and I wrestled with it in a course I taught for many years before launching the class I teach now, so I had many chances to be challenged by Baldwin.

One of my favorite passages in the text has one of the protagonists, Vivaldo, on a West Village rooftop at night with friends. He’s resting on his back, knees up, looking at the sky overhead. He sees it this way:

The sky looked, now, like a vast and friendly ocean, in which drowning was forbidden, and the stars seemed stationed there, like beacons. To what country did this ocean lead? for oceans always led to some great good place: hence, sailors, missionaries, saints, and Americans.

What a paragraph!

After reading the full scene out loud in class together, a student or I would first draw a stick-figure diagram on the board so students could see the scene Baldwin was laying out. We would then start annotating that picture, drawing arrows to each person and symbol and jotting down what they seemed to mean.

Then we got into the language.

Baldwin has the sky become a “vast and friendly ocean.” The stars become “beacons.” The ocean-sky leads to some “country.” (This is where at least one student each period would flip back to the cover of the novel in excitement, remembering the book is titled Another Country.) Baldwin inverts the world, with Vivaldo looking down at the ocean-sky beneath him, sticking (like Spiderman, some student would often point out) to the upside-down rooftop above.

That’s usually all we got through before it was time for students to write and reflect on Baldwin’s argument by themselves, then discuss as a whole class, and then write some final takeaways before the bell rang.

Looking at the passage now, though, I see some ideas we didn’t get to. Maybe that’s for the best — those other ideas were meant more for me than for a classroom full of young people of color. Or maybe that’s all the more reason for us to have seen them? What would Baldwin have wanted, I wonder.

Let’s see.

Vivaldo is white, in a novel in which characters’ race features prominently. I’m white. Vivaldo is looking at an ocean that is “vast and friendly” and leads to “some great good place.” Baldwin’s suggesting, in part, that this is the ocean that led to America. And here is a white American who sees friendliness and goodness — and greatness — reflected back at him when he thinks about the journey to this country. There were other journeys to this country on that ocean, though.

Is that why Baldwin ends the paragraph with “hence, sailors, missionaries, saints, and Americans”? What is he saying, then, about Americans? About us? About me? What do I see reflected when I look at his word-waves?

Here’s what AI sees.

I just put this prompt into Fable, Anthropic’s most advanced Claude model:

It “thought” for about two minutes. Then it came back with this:

This passage is Baldwin working in his signature double register: it begins as consolation and ends as a joke with teeth, and the interpretation lives in that turn.

Start with the inversion. The sky becomes an ocean — above becomes below, air becomes water…

Then, a little later:

Since oceans always lead to some great good place, whole categories of humanity are explained by the crossing: sailors, missionaries, saints... and then “Americans” lands as the punchline. A nation, Baldwin implies, constituted by a faith about destinations: Puritans, the city on a hill, the immigrant’s golden shore… But the wit is also an indictment, because the sentence conspicuously omits the one Atlantic crossing that founded Black America.

It concludes:

So the passage performs, in three sentences, the whole novel’s argument: it grants the beauty of the hope and then gently exposes it as the most American of illusions, the story people tell about the horizon. Baldwin never quite renounces the hope. He only insists we notice that the horizon, once reached, is never the great good place. It is only ever another country.

I have to admit, I’m impressed. I’ve spent most of my adult life, first in college as an English major and then in education as an English teacher, analyzing texts carefully. In two minutes, Fable hit every major note students and I ever did over many years of conversations about this passage. It also explored the implications above that I didn’t take our classroom into, for better or worse.

But Fable showed some of its cards as it was “thinking.” Here’s one moment while it gathered its “thoughts” before responding to my prompt:

It was reading what other people had to say about all this. Or more accurately, it was scraping up some search data so it could better predict the next token as it answered my question. Notice it didn’t reference The Bosphorus Review of Books or Goodreads in its answer. All those ideas were presented as Fable’s, and implicitly as original. This is all you see when the response is done:

You can click on the line at the top, “Excavated literary layers…,” and see all the “thinking” again. But that’s not what the AI’s design is encouraging. Why not, a reader might ask. What’s the subtext?

Now let’s get into the language:

This passage is Baldwin working in his signature double register…

“Signature double register” is exactly how Baldwin’s voice shouldn’t be described. “Signature” as an adjective is a branding term — signature scent, signature dish — while I associate Baldwin’s work with an interrogation of signatures and signing, the noun and verb.

And the closing continues the classic Claude tone:

He only insists we notice that the horizon, once reached, is never the great good place. It is only ever another country.

In my experience, Claude has been designed to go for a polite clincher. Politeness is another kind of prose Baldwin rejects. His words — especially last words — ring, trying to wake us up, not to soothe us with a false finality. Baldwin likely would also have had some thoughts about the relationship between Claude’s designers’ experience of America, Claude’s training data, and that signature Claude politeness.

Ironically, I suspect that if Fable had simply read Another Country when it responded to me, instead of reading about it, it might have answered in a way that was more true to the novel’s spirit. But that’s part of the point: AI is agglomerating and synthesizing, not focusing.

Fundamentally, Fable’s prose exposes how little it has at stake. When I read that passage and come to those same ideas, I feel uncomfortable. It’s hard to keep looking at them if I’m really looking at them. Those are deep waters.

But that same analysis came very comfortably to Fable once it finished deciding how to combine training data, search results, model weights, probability, and inferred user preference, and began unspooling its tokens. Fable could do this all day.

Seeing that difference is what it means to read AI. So is opening the hood and looking at how Fable came to those ideas in the first place.

AI literacy means, in part, teaching students to read AI that way. To realize that the warning underneath every prompt window should actually say, “AI can make mistakes, but drowning has been forbidden.”

I’m sharing my current thoughts in case they’re useful to others wrestling with similar questions in their contexts. In this post and some future ones I’ll engage with key questions I think AI literacy advocates need to answer. Today’s question, to start this off:

What does the phrase “AI literacy” even mean?

When I use the term, I’m thinking about the literacy I explored with students for many years as an English teacher. I understood my primary goal to be teaching reading, writing and discussing. By the end of high school, this mostly meant helping students solve a kind of puzzle, albeit one that can be assembled many different ways.

Teaching reading meant helping students put together the pieces of a story, the events of the narrative and the words the author uses to describe them, so we could see the broader picture the author was creating — and so students could later do that again on their own. Teaching writing and discussing meant helping students find their own words for the ideas about a text or about life (or ideally about both) that they wanted to convey. In other words, students would write and talk about the process of putting together the puzzle of a book and what they learned about the book and themselves during that process.

Some of my most powerful literary experiences, as a teacher and a reader, were working through the novels Another Country by James Baldwin and The Women of Brewster Place by Gloria Naylor in a 12th grade English course I designed and taught for several years. In both books, the authors weave together symbols and images to help readers confront truths about America and about themselves.

Baldwin’s characters listen to and perform classic blues songs throughout his book, and we would play the original audio in class, annotating the lyrics and listening to the melodies and asking why Baldwin would put that song in this story at that time. Then we’d consider whether we might still need that kind of music — literally or figuratively — in our time and if so, how to make it. Naylor uses the idea and image of a phoenix to launch her book, so in each character’s chapter we looked for the moment a phoenix would appear, sometimes as a reference to sunlight and sometimes as a pigeon taking flight. We’d then discuss the implications of Naylor’s narrative, whether we should be looking for phoenixes in our lives, around us and inside us, and if so how to go about finding them.

In both cases, learning to read, write and discuss better also meant learning to live better, and live together better — the text providing a kind of secular spiritual training for both the individual student and the classroom as a whole.

When I think about AI literacy, the translation I realize I’m making is to treat AI as a text. The technology of books, as it were, is something we barely notice any more. But when I think about reading, I’m thinking about engaging with structures and conventions that have become almost subconscious in their ambience. When we read Naylor and Baldwin, we would flip between different chapters to see the threads the authors tied to hold their story together. We would zoom in on specific words and phrases the authors put down on the page. We’d ask, over and over again, “Why did the author make this decision?” We’d consider the world they were writing within, what they thought of that world and what they wanted to change, and also consider the world we were reading them within and what we might want to change.

In English class, the book creates an occasion for us to work together to understand ourselves, each other and our moment, and then ask what should happen next. In AI literacy lessons, I understand AI to be doing something similar.

A key move in both cases is to see the “text” as full of human decisions.

In English class, that aspect of literacy is easier to convey. The author’s name is on the cover. The publication date is given a few pages before the story even starts. We know right away who made this and when, so we can move on to the question of why.

In AI literacy lessons, we need to help students see the authors. For example, as we redesign our AI literacy class for 26-27, Scott Kern and I want to spend more time introducing students to the leaders and researchers shaping AI systems. We want to give students a sense of who those people are, how they think, and what their formative life experiences seem to have been, in the same way I’d always devote a lesson or two in that English class to background on Baldwin and Naylor, including videos of interviews with them. In both cases, the goal is to help students see that there are people who made decisions about the experience they are having, about how the “text” in front of them was built.

(It’s also important to remember that authorship is a more complicated concept when it comes to AI. The incentive structure surrounding this technology is an author too, as are the many people involved in training AI models, as are we in the form of the training data we collectively created.)

Once we make the decision makers visible, we need to teach how to identify their decisions and consider the implications of those decisions. Baldwin keeps alluding to the blues — what does that do to our experience as readers, and why do you think he decided to do that? Naylor keeps alluding to phoenixes — what does that do to our experience as readers, and why do you think she decided to do that? AI leaders and their UX teams designed AI conversations so that they look a lot like a text message exchange — what does that do to our experience as users, and why do you think they decided to do that?

Part of the point is to help students wake from the dream we can all fall into when reading, or when using AI. In both cases, we can quickly forget we are living in a world someone else created and start to feel like things were always going to be this way. Another Country was inevitably going to be full of music; LLMs were always going to respond using the first person. When we remember the authors, though, we wake ourselves up, which is to say we remember who was doing the dreaming in the first place. (Obligatory former English teacher note: The narrative dream is a wonderful one! I just wanted students to know how to make it a lucid dream from time to time.)

A second key move is to see ourselves as decision makers too.

That’s why our English classes ideally landed on questions like, “If Naylor designed her book to make an argument about the importance of personal rebirth, what would it mean for you to rise from your own ashes? What would it mean for us to collectively do the same?” The point was for students to read books and write about them until they realized they could also read and revise reality.

In AI literacy lessons, the questions look like this: “If AI companies designed LLMs so it seems like you’re talking with an all-knowing, infinitely patient conversation partner, not just prompting a probability machine, what would it mean to use AI in a more lucid way? What would it mean for us to collectively argue for different AI design choices?”

That’s why I think the term “literacy” works for the kinds of AI lessons we should be teaching. In the way I understand it, “literacy” means practiced awareness and analysis that leads to increased agency.

In both kinds of classes, though, the literacy edifice rests on a potentially fragile foundation.

When I designed that 12th grade English course, I built in mechanisms to hopefully ensure students did the reading before class. We had “reading tests” on Mondays, where students answered questions about the plot of the weekend’s reading (and could use notes to do so, which they then had to turn in with their test). But in addition to extrinsic measures like that, I also hoped our classroom conversations were so intrinsically interesting they became another motivation for doing the homework. The larger point here, as any English teacher will tell you, is that we need everyone to do the reading.

And that’s the most basic reason I believe we need to bring AI into high schools, in the same carefully and professionally curated ways that we bring in novels like Another Country and The Women of Brewster Place. Both were explosive texts in their time and still are in ours, based on the way students carried our conversations out into the hallway after class. AI is a new kind of text we need to help our students engage with, one that is powerful, exciting and dangerous. AI raises new — and old — questions students need to grapple with, and their teachers are some of the people best equipped to help them do that. But they can’t do that grappling, and we can’t help them do it, if we leave them alone to encounter AI at their bus stop, on their couch, or in their bedroom. We all need to be doing this reading together.

We walked the girls up the steep hill from the campsite to the train tracks, sliding on loose dirt and grabbing young trees for leverage. We’d already told them the story of our first time camping there in the late 90s, when we didn’t know the tracks were there and were woken suddenly in the night by the roar of a passing train. I can still remember its light filling my tent and its sound seeming even louder in my half-awake confusion. Luckily there were no nighttime trains along the Delaware that weekend.

Trains have been on my mind recently, which meant Thoreau wasn’t far behind. I first read Walden in college and at the time thought Thoreau to be the height of profundity, partly because of his thoughts on trains. I later decided he was more obnoxious than profound — his mother was doing his laundry for him that whole time! — but have now settled somewhere in between those poles.

In this moment, his reflections on trains feel like a good place to turn as we confront the AI models racing by us in the news and in daily life. Thoreau was writing Walden soon after the Fitchburg Line first started to run by Walden Pond, and he was a teenager at Harvard when the first New England train line was introduced. He was growing up amid the start of a significant technological and social change, and spent time and words to explore its implications.

In the chapter “Sounds,” he writes:

As I sit at my window this summer afternoon, hawks are circling about my clearing; the tantivy of wild pigeons…gives a voice to the air…and for the last half hour I have heard the rattle of railroad cars…For I did not live so out of the world as that boy who…[said] you couldn’t even hear the whistle! I doubt if there is such a place in Massachusetts now.

I doubt there is a place in the US now where the phrase “AI” or “that’s AI” or “that’s not AI” can’t be regularly heard. Even those not spending too much time thinking about all this are likely to encounter the notion of AI on a near-daily basis. In my anecdotal experience in both school and the rest of my life, young people increasingly say “that’s AI” to mean something like “that’s gotta be fake” or “you’re kidding me.”

Thoreau’s opinion of trains is mixed. On one hand, later in that same section he compares the railroad to natural phenomena whose power and beauty he clearly appreciates:

The whistle of the locomotive penetrates my woods summer and winter, sounding like the scream of a hawk sailing over some farmer’s yard.

The engine with its train of cars moving off with planetary motion,—or, rather, like a comet…since its orbit does not look like a returning curve.

I watch the passage of the morning cars with the same feeling that I do the rising of the sun, which is hardly more regular.

A hawk, a comet, the rising sun — on one level Thoreau was impressed with the capability of this technology, and with the economic momentum it both embodied and accelerated. He continues:

What recommends commerce to me is its enterprise and bravery. I see these men every day go about their business with more or less courage and content…On this morning of the Great Snow, perchance, which is still raging and chilling men’s blood, I hear the muffled tone of their engine bell from out the fog bank of their chilled breath, which announces that the cars are coming, without long delay, notwithstanding the veto of a New England north-east snow storm.

Trains felt inevitable to Thoreau, as both a fact of daily life and a fact of that historical period.

But their physical inevitability did not mean their implications were equally inevitable. Here he is two chapters earlier:

It lives too fast. Men think that it is essential that the Nation have commerce, and export ice, and talk through a telegraph, and ride thirty miles an hour, without a doubt, whether they do or not…If we do not get out sleepers, and forge rails, and devote days and nights to the work, but go to tinkering upon our lives to improve them, who will build railroads?

This is one of those times when, for me, Thoreau’s insight exceeds his sometime-insufferableness.

“If we…go to tinkering upon our lives to improve them.”

I went for a walk in the Maine woods this morning with my daughters and their cousins. At one point my younger daughter sat down by a small tree stump, criss-cross-apple-sauce like the first grader she is, and started to decorate it with nearby leaves, sticks and pinecones. The rest of us chatted and then, when it was time to keep going, my daughter asked me to take a picture of her handiwork.

Later on, after that walk was over, I was sitting and looking out a window at that same forest when I found out Fable was returning. The latest model of Claude would soon be back, now that the government had rescinded its recent de facto ban.

I was excited to again try out the most capable AI model yet and see what its implications are for my work and my students’ and daughters’ futures. I think most people should use Fable at least once to support a goal of theirs, just to see what these systems can now do. Among other things, I’ll use Fable to get feedback on this piece and to help me think through the best way to approach AI literacy with high schoolers this fall.

Looking at that moment through Thoreau’s eyes, though, I see Fable — and AI more generally — like a train barreling through the woods. I heard about Fable’s impending return before it was actually accessible on my account, like “the muffled tone of [the] engine bell…which announces that the cars are coming, without long delay,” notwithstanding the veto of a Washington DC political storm.

More generally, the new AI models just keep coming, leaving those of us in the AI-in-education world often feeling like Thoreau “watch[ing] the passage of the morning cars with the same feeling that I do the rising of the sun, which is hardly more regular.” AI releases don’t have that kind of chronological regularity, but they are fast becoming regular in the sense that “sure as the sun will rise” could be restated as “sure as the next model will get shipped.”

Thoreau ends the section in “Sounds” with this:

I will not have my eyes put out and my ears spoiled by its smoke and steam and hissing.

As much as trains looked and sounded to him earlier like near-magnificent forces of nature, he closes with their “smoke and steam and hissing.” They remain forces of nature, but now of an economic nature, possessing a power that needs to be both leveraged and reckoned with. And the point is our eyes and ears.

Here’s the stump my daughter decorated:

It’s not until putting it here, now, that I see the care with which she placed everything. She was humming quietly to herself as she worked. That, and her family’s conversation nearby, and a mild breeze swaying the pine trees were the only sounds in the forest.

Which do you prefer, the boat or the train? Which do you want for your kids?

What makes us think we — or they — have to choose?

We aren’t the first people to face a technological and social moment like this one, full of sudden, seemingly binary choices that require imperfect decisions. Something similar was happening with railroads in England in 1844, and the way one painter reflected on his time can help us better engage with ours and the choices we now need to make.

A year ago, before I launched this Substack, I took a first look at this painting in a LinkedIn post to see what it could help me understand about our AI moment. I wanted to look again now, a year later, to see what I might have missed.

Like last time, the first thing I notice is the sheets of rain, but the haze they make, which a year ago seemed ready to swallow the rest of the painting, now appears thinner to me. I can see a clarity beyond it, though not an easy one.

The next thing I see is where Turner positions us, the viewer. I hadn’t realized before how much we are in the picture. But there we are, in the middle of the frame, looking at the hazy center where the bridge on the left and the train tracks converge in the distance. The train will speed by on our right, roaring in our ears, forcing us to flinch, and maybe even knocking us off our feet.

I find myself resisting the train, knowing I need to look at it but wanting to look elsewhere first. Its commanding, stark profile, shown with the darkest colors in the painting, is off-putting in its self-assurance.

Better to leave the haze and head to the clearer left, to alight on the river and its boat.

It’s a calm scene, both in its movement and its colors. The two passengers, lit maybe by setting sunlight through rain, are resting, or at least moving along slowly. The green of their sail (parasol?) echoes back from the still water.

I’m taken back to 1999, when I started college, and the small lake on Stanford’s campus was full of water. It turned out to be full rarely, so this was a lucky break. Some enterprising soul had made a makeshift raft, and new freshman friends and I would go out on it, pushing off from shore with a long stick not unlike the one above. We spent many sunny, aimless afternoons on the lake together.

People all around us, including many we went to classes alongside, were busily building the search engines and apps and devices that laid the groundwork for the online world we now live in and the AI one approaching. I still remember the first time I heard the phrase “natural language processing” explained by an acquaintance.

But sitting or lying on the raft, all you could hear was a breeze and the flapping of ducks’ wings and muted conversations as people walked on the shore. When there was movement, the main form of propulsion we were aware of was the wind, which kept stubbornly pushing us into clumps of reeds we’d need to push our way back out from.

Looking back on it, though, a train was leaving a station nearby, propelled by visions and incentives both timeless and very of that time, which is to say of this one.

Back to Turner’s train.

There seem to be passengers on it, too, but they are more numerous and crowded and agitated (and delighted?) than the two on the boat.

But they are visual afterthoughts to the locomotive and its fire. The engine’s glow threatens to engulf the train, as if its own momentum might consume it, or us. Blake’s “dark Satanic Mills” come to mind.

Looking at the whole image again, I now see a vertical line down the middle to match the horizontal one that divides the land and sky. The vertical border divides — and therefore unites — the boat on its river and the train on its tracks.

They are both going diagonally, and the diagonals converge at a vanishing point lost in the haze, but their current trajectories are pulling them apart. These are two different vehicles with two different destinations.

Their colors also pull them apart while stitching them together. The light hues of the left and the dark tones of the right merge in the middle.

It’s tempting to see the left side, with its human-powered boat near what almost looks like a Roman aqueduct, as a totally different era, one quickly being left behind by the modernity on the right, with its coal-powered train and newly engineered bridge. But Turner put them together on one canvas, and left them there for us to reckon with.

The train will pass us first, momentarily. Will we get on or be run over? Or will we be forced to leap and fall into the river below?

Let’s imagine we keep our footing as the train roars by. The boat will eventually make its way by too. Do we drop down to board, or let it drift past?

But Turner didn’t really leave us space to get on either. The boat and train are full. And they won’t pass us. Movies hadn’t been invented yet, and even if they had been that’s not the game Turner was playing.

He leaves us looking at both boat and train, so that we have to reconcile them. Which means reconciling ourselves to a present where the future and its implications rush at us while the past and its reminders remain in sight. And which means making our own decisions about the colors we will paint each of those in, how we will balance their sizes, and ultimately what kind of canvas we choose to create.

Seen that way, this painting is less an argument than an invitation. Welcome to another time it felt this way. Out of his time Turner made an artwork that looks back at us to ask: What will you make of yours?

The decisions we make will become the canvas our students and children step into. My older daughter is already halfway to college, where new boats and trains will be waiting for her and her friends to make sense of. The next time we’re out at Stanford, I need to take her out to the lake — which I think is mostly dry these days — to help her see that the point of a raft is the other people on it.

One of the reasons I first started exploring AI in education was to prepare my students for the inevitability of the AI future. That’s not my motivation any more.

After ChatGPT came out in 2022 and each new model followed in a steady drumbeat, my initial teacher instincts told me it was time to prepare students for a new world. The program I run for 11th and 12th graders has always had preparation for life after high school as its goal, and of course education more broadly is always a kind of preparation. So AI, and the changes to the world it promised, became for me another thing to prepare students for.

The first two school years after ChatGPT I learned how to use LLMs myself, wove them into my teacher workflow, then brought custom bots into my classroom. The summer after that, alongside Scott Kern, I co-developed an AI literacy framework for high schoolers — The AI Driver’s License — and then co-taught a pilot class based on that framework this past spring.

All the while the pace of AI development has continued to accelerate and the conversation about that development has become ever more breathless. This summer, recursive self-improvement and world models appear like seabirds before landfall, heralding a new world of AGI over the horizon, the currents of history carrying us inexorably toward its shores.

At least that’s how the story goes.

As I look ahead to next school year, I haven’t changed my feelings about the importance of AI literacy for high school students, especially a literacy that helps young people become critical users who can treat AI as both a tool and a topic. I’ll keep exploring new ways to use AI in my work and to bring it into my classroom.

But I’m developing a new motivation for my AI work. I originally wanted to prepare students for the future; I now want to prepare them for futures.

The AI discourse of the past few years has created a mental habit I’m only now starting to see clearly in myself. It’s the habit of thinking that there is one future we’re all careening towards. And more than that, it’s the habit of thinking of that future as being created by others, whether those others are the builders of AI or, increasingly, AI itself. From this perspective, it’s no wonder some AI leaders use the word “singularity.” In the story they tell, AI is a singular technology created by singular minds in singular labs, supported by singular amounts of money, which will have a singular impact on a single future.

All of which means that if we as educators try to prepare students for “the future” with AI in mind, we are surrendering to a specific narrative. And there is another surrender under that first one. If we’re not careful, AI literacy becomes a form of passive acceptance, a way to normalize AI. The logic would go something like this: “Writing and books are a fact of life, so we need traditional literacy. The internet is a fact of life, so we need digital literacy. AI is a fact of life, so we need AI literacy.”

The original literacy — the written word, books, etc. — is settled fact. But even with the internet, the facts are clearly not settled. Thirty years since I listened to my modem log me onto AOL, we are still litigating, figuratively and literally, the shape of the online world and its effects on the offline world. It’s only been three years since ChatGPT. The AI world has barely begun; many possible futures are on the table, or at least should be, and we should be debating them as such. If we fail to do so, if we treat the AI future as singular, not only do we fail to prepare our students effectively, we end up carrying water for those with the most to gain from that singularity.

The real preparation lies in helping students see that the future is not a world others are building that they need to find their place within. Just because a company can build a data center does not mean it can create the world for that data center to exist in. Our current data center debates are proof. People with many perspectives and motivations are raising their voices over how or whether construction should proceed. That’s not what inevitability sounds like.

There is no future yet. There are many futures flickering in the distance, and students need to decide which one they want to live inside and then start building it together. That’s what I now mean by critical AI literacy.

I do believe that the advent of AI is a historical event, but like all historical events its meaning will be determined first by the people living through it, and our students will be living through this the longest. They have a right to self-determination, now more than ever, and we have a duty to teach them what that means.

We ended our AI literacy pilot this past spring by having students make “AI declarations” modeled on the second paragraph of the Declaration of Independence. First they wrote their own paragraphs starting with “I hold these truths to be self-evident.” Then we came together as a class to synthesize everyone’s individual beliefs about AI into a collective document that began “We hold these truths to be self-evident.” We worked through the wording of that collective AI declaration together in our final class. Here’s an abridged version of where we all landed:

First, AI is a tool, not an intelligence. It’s a talking library built from everything humans have ever known.

Second, AI should sharpen human thinking, not replace it. Learning is lost when convenience becomes more important than growth.

Third, when AI begins to negatively affect our creativity, our careers, our environment, or our connection with each other, we have the right and the responsibility to question it, limit it, and demand it be regulated for the good of all.

We originally came to the declaration activity inspired by America’s 250th anniversary and hoping to help students see that they can make declarations in the midst of historical change, as the Declaration’s authors did before them.

I now increasingly see how important the independence part is too. To teach our students about AI, we need to teach them how to be independent thinkers who can free themselves from the tunnel vision constraining the AI conversation. That will require us to free ourselves first, to realize that we are not preparing students for the AI future. We are helping them see there are many AI futures possible and they can decide which ones to fight for.

I had just walked into the gym a little late, having been fetched by a colleague because I had gotten too engrossed in a national AI-in-education call in my classroom.

Our principal got everyone’s attention and the rehearsal began. Students moved from the bleachers to their assigned seats on the basketball court, arranged as they would be on stage. Their teachers helped them find their place.

We then walked out of the gym in two long lines, students in the order they would be in when they walked into the theater for the real thing. Our school doesn’t have enough hallway space on the first floor to fit everyone, so we headed up the stairs. We had to go up five flights the first time, teachers huffing and puffing along with students.

As we waited, students waved to others further down the line. Those fortunate enough to be alphabet neighbors with best friends chatted quietly. Teachers joked about fitness at various stages of life and exchanged knowing remarks with students they had spent every day with since August.

The seniors started back into the gym, practicing their entrance to “Pomp and Circumstance.” As I watched students on lower floors walk by, I had flashbacks to moments involving them in my classroom this year. In the midst of making their own memories, students were already partly memories too.

Once we had assembled in the gym again, students in their seats and teachers standing nearby, we rehearsed the rest of the program. Teachers introduced our salutatorian and valedictorian, who each came to the podium and said “My speech” instead of delivering it. The school choir sang the start of the Black National Anthem and our alma mater.

Two seniors, Nyla and Yasmina (pseudonyms as always), then rose and did an a cappella run-through of the song they’ll perform at graduation, “A Million Dreams” from The Greatest Showman. For the first time all rehearsal, the gym fell silent. When they finished there was loud applause.

After that, our principal led seniors through the call-and-response that accompanied them through elementary and middle school.

Who are you?

A star! I shine brightly for others!

And why are you here?

To get an education!

And why else?

To be the great person I am meant to be!

And so on. This is not the kind of thing teenagers will entertain much outside of moments like this, so several had forgotten it since they last recited it in 8th grade and had to review printed copies of their lines. I stood to the side reminding those nearby how it went; I had joined students in this call-and-response many times in my first five years when I taught at our middle school. Eventually they all had it down.

The principal ended with the line that closes graduation:

Seniors, turn your tassels!

Many turned imaginary tassels or real baseball hats. A few threw invisible mortarboard hats into the air and cheered.

Then it was over. Students made their way out, some stopping to say hi to me and other teachers. We wouldn’t see them again until their last day of high school. They looked a little dazed, like they had just walked out into sunlight forecast but not yet felt.

I started to realize that I had spent much of the rehearsal trying to figure out how all that related to AI. I had run through a few options while watching everything unfold:

• Rehearsal for real life requires being in real community.

• Imaginary tassels, real pride — no screens necessary.

• Getting ready for the new career ladder by walking to the fifth floor.

But as I left the gym and headed back to my classroom, where I can still see everyone at their seats if I close my eyes, I realized the point was not to insert the afternoon into an AI narrative. It was to protect it from that narrative.

Students like Jessie, Veronica, and Russell were at the center of many memorable AI-related stories this year. And I’m already excited for the AI stories to come this fall, as the new seniors and I continue to explore this moment together.

But this week it was time for this year’s seniors to be themselves, not just students in the AI era.

I sat down at my desk, the walls behind me newly bare as we pack up our posters for the summer. I wondered, after writing about AI in education weekly for a year and thinking about it daily for longer than that, if I still knew how to be myself, not just a teacher in the AI era.

Can I still tell another story?

In his set “The Dreamer,” Dave Chappelle tells a story that implies a question I understand better now, as Claude Fable drops to fanfare and the clamor about AI IPOs and AI tokens and AI backlash and AI hype and AI slop and AI everything gets ever more deafening.

Whose dream are you living in?

When Nyla and Yasmina sing at graduation, I’ll film them so my first grade and fourth grade daughters can see their performance, so we can see what dreams should be made of.

On our school’s annual Community Service Day, we send as many students as we can out into the world to plant gardens, play bingo with older residents, clean up parks, and generally be of service. But we couldn’t connect with quite enough external organizations for every advisory to be of service out in the world this year, so some stayed at school. Several teachers came up with service-oriented, hour-long sessions for them to rotate among. I was asked to do one about AI.

I’ll share here what I did for that hour, because by the third time through I was realizing there might be some useful ideas for other educators in the session. I also thought my uncertainty by the end might be helpful too, since I imagine that’s what many of us feel as we feel our way through the AI moment.

Minutes 0-5

This slide (made with Ideogram) was up when students walked into the room:

I opened with the line from the start of this post. When students nodded, I added:

AI is here. It’s going to affect the world you graduate into in a few years. You’re doing a service to our community by knowing what it can do and also what the implications are of what it can do. My generation is mostly creating this opportunity/problem, so we need your generation to start thinking about how you want to handle it all. That’s what we’ll be doing for the next hour: talking about how you think we should handle AI.

But first, you’re going to make some songs.

I then explained our opening activity.

Minutes 5-15

I switched to this slide:

Students worked in pairs or small groups to come up with the most original or off-the-wall artist collaborations they could think of, singing about whatever topic came to mind. Two of my favorites over the course of the three sessions:

The Weeknd and Beethoven: “Five Minutes,” a song about only having five minutes to find your homework before leaving for school.

Kendrick Lamar and Olivia Dean: “Half-Empty Tank,” a song about musing on recent heartbreak while driving an expensive foreign car.

To make all this happen in real time, I had students tell me the artists and song topic. I then gave that information to a Claude project I had prepped to churn out song titles, lyrics, and style notes based on the artists and topics we put in. I then cut and pasted those topics into the relevant boxes on Suno and let that AI make the music.

As the first songs came on, students were taken aback. It seemed many hadn’t realized how far AI music generators had come. They were nodding along to the beats and saying “oooh” at lyrics like “I gave you my heart, but you wanted the bank / Now I’m in the coupe with a half-empty tank.”

But after the initial shock wore off, they were less and less impressed. They started pointing out that the voices didn’t sound like the artists we had put in. I explained the copyright issues. They complained that all the songs were starting to sound the same. I explained the model was probably tuning itself to our requests during this session.

When it was clear they had heard enough, we started a short discussion about the whole experience.

Minutes 15-25

To spark the conversation and give them a little more context, we started by watching parts of this video about Suno’s business model and the mixed artist reactions to it. (We watched 0:00-1:35 and 2:45-3:00 for time but the whole video is useful.)

I then asked students: “What do you think about AI music? Should it be allowed? Will there be a time, before you graduate high school, when people will develop preferences for AI music versus human music? Will some people say things like ‘I turn the AI music preference button to off when I stream’ in the future?”

Students talked in groups. They were mostly of the opinion that AI music is a negative development. They said things like:

“Human artists need to be the ones creating music. We have human imagination.”

“AI hasn’t had any experiences. How is it supposed to sing like it’s feeling something?”

“Real music takes time.”

To end the conversation, I said, “So you think you’ll always be able to tell if a song was AI created?” Most of them nodded, but they were a little less sure about that. I then said, “Let’s see if you can tell when a movie is AI created.”

Minutes 25-40

I explained briefly what Seedance 2 is, and then we watched those clips of Tom Cruise and Brad Pitt fighting that had caused such a stir in February.

We discussed if students could tell that was AI. Some said it was obvious, others were not so sure. But most students believed they would be able to tell if a movie was AI.

I said, “Ok, you all seem pretty confident about this. Here’s a short film, just two minutes long. It was made by someone named Sebastian Lopez. But did he use AI or really film this? At the end I want you to tell me if this is AI or real.”

We all watched. At the end, students talked in groups. We then took a vote. In each period it was basically fifty-fifty between AI and real. (My guess is this was partly the point Lopez was trying to make in the clever choices he made as a director.)

I then revealed that the film was a genuine human production, with no AI involved as far as I knew, except maybe that part with the six-fingered hand.

I asked: “If you all, young people who watch a lot of movies and shows, are split 50-50 on whether something is AI now, and it’s only 2026, where is this going? Will there eventually be AI movies you don’t know are AI? Like with music, will people start to develop preferences for AI movies or human movies, like each is its own genre?”

We had a spirited, though too short, conversation. It was time to pull back the curtain on TikTok.

Minutes 40-50

We’re going to play a game using TikTok (or another social media like Instagram if you prefer). Here’s what you need to keep in mind: these platforms want to keep you watching. That’s their goal because it’s part of their business model. They make money from your attention. So their goal is to get you to keep watching once you’re on.

And to do that, they use AI. There is an AI choosing the videos you watch on TikTok or the pictures you see on Instagram.

So…what is the video you think TikTok’s AI is going to show you when you open it up in a moment? Don’t open it yet! Think about it: what video do you think TikTok’s AI thinks is going to hook you right away? What topic or person or situation is the AI going to choose for you?

Students shared their guesses in groups. They then opened their apps to see what the AI chose.

Their reactions were a range; some students seemed to just get lost in the TikTok stream immediately, while others maintained enough critical distance to see the choice the AI made. Those students woke their friends up from the social media trance that had set in, saying:

“I was just talking about buying this game!”

“I was literally just thinking about what tattoo I should get.”

“I’m always hungry at this time on Fridays and they started with an Applebees ad.”

I’ve heard these conversations devolve into conspiracy theories about TikTok secretly turning on the mic on people’s phones, so I jumped in quickly:

This is not a conspiracy. It’s something you agreed to. We all do. It’s what’s in those terms of service we never read and just agree to when we download an app. I do the same thing with LinkedIn, the boring social media I’m on. Now I’m going to show you what the AI inside social media does with the information you agree to give it.

We then watched three minutes of The Social Dilemma, from 16:30 to 19:30, pausing in the middle to linger on this shot:

I read off the lists on both sides of the screen and said,

This isn’t a real screen inside TikTok; it’s a dramatization by the documentary makers. But TikTok and Instagram do know these kinds of things about you. And you don’t have to tell them any of this for them to know it. They’re deducing it, using AI, based on which videos you watch, how long you watch them for, which ones you like, etc.

It was at this point in the second session of the day that one freshman’s jaw dropped. Having taught English for many years, I mostly associated this expression with an extremely dramatic plot twist in a story. I hadn’t seen it in an AI context yet.

The other thing I hadn’t seen quite as clearly was the fear he was suddenly feeling. He was sitting near the window, and the light coming into the room illuminated his widening eyes. And as any teacher will tell you, when one student is visibly feeling or thinking something, many others are too, just less visibly.

I suddenly wondered if this session, taught to students I don’t usually teach, had been both too effective and not effective enough. In classrooms, pacing is often the secret to a good lesson. But the unforgiving pace of the AI era can make pacing in AI lessons hard. I had five minutes to land a plane I’d flown into turbulence of my own making.

Minutes 50-55

Remember what I said at the start? It's literally a service to our community for your generation to be thinking about all this. You're the ones inheriting this world, AI parts included, and we need you thinking critically about this tech, not just using it. But the thinking you're doing is also power. Next time you open TikTok, you'll be more aware of what its AI is trying to do to you. Now you can watch TikTok, not just watch TikTok. You can break the spell. Same with AI music or movies. Once you know how it works, you can decide what we should do about it.

You see what I mean?

I’m not sure they all did. With class rapidly ending, this was not my best closing or checking-for-understanding moment.

In my experience, timing is a central tension in all AI work with students these days. If we give students enough time to really see what AI can do, there often isn’t enough time to reflect and debrief on the implications. If we make time for the discussion, it’s often too thin because it’s not based on an understanding of AI’s current capabilities and future trendlines.

To put it another way, my AI work with students often see-saws between emotions: excitement and confusion, intrigue and avoidance, curiosity and fear. One class period never seems to be enough to find balance. And unlike teaching English, when I always knew how the story would end, this time I’m in media res with the students. It’s hard to plan for plot twists.

But despite my regrets about this particular lesson plan — too many activities in one hour, too little time to discuss — I still think it was important to try. To me the real mistake would not be a poorly paced lesson that mistakenly leaves students with more fear than agency. It would be letting students continue to use TikTok like no one’s watching, or letting them go through high school like AI isn’t waiting outside.

I didn’t subject the underclassmen to TS Eliot, but this part of “Four Quartets” was on my mind as I wrote this a few days later, feeling like a veteran educator who’s somehow a first-year teacher again when it comes to teaching AI:

Trying to use words, and every attempt
Is a wholly new start, and a different kind of failure
...
but there is no competition—
There is only the fight to recover what has been lost
And found and lost again and again: and now, under conditions
That seem unpropitious. But perhaps neither gain nor loss.
For us, there is only the trying. 

Minutes 55-60

As they walked out the door, I played some of the Suno songs from the start we hadn’t gotten to. One student asked to hear his K-pop/techno brainrot ballad “Ohio Glitchgyatt” again, which I understood even less the second time. I heard another student joke to a friend, “We should’ve asked people to stop using water while we were making all those songs.” That felt like something the students out gardening would appreciate. Then the next group walked in.

I was eager to get to the section about schools, but partly because I was now on a highway and partly because I was interested in how the pope would set things up, I let it play from the start. I’m glad I did — I might have missed the most important part otherwise.

The first two chapters of the document focus largely on the “Social Doctrine” of the church. I’m not a Catholic or religious in general, so I was interested in the concept not as theology but pedagogy. And one particular subsection of the Social Doctrine discussion had pedagogical implications to spare.

With only six days of classes left this year (!), I don’t have time to explore these ideas with students now, but I’m already starting to imagine lesson plans for the fall in the purpose development and career-connected learning program I run.

The Common Good, the For You Page, and the Lunch Spot

Chapter Two of the encyclical explores the principles underlying the Social Doctrine, and the first it names is the principle of the common good. The Pope starts his explanation this way:

The common good consists in the sum total of social conditions which allow people, either as groups or as individuals, to reach their fulfillment more fully and more easily…It is not the sum total of individual benefits, nor the intersection of their particular interests; it is a greater good that belongs to everyone, and it can only be achieved, nurtured and protected by our collective efforts. (Paragraph 60)

I’m going to have students read and annotate that paragraph, first with literal paraphrases of the Pope’s meaning and then a second time with their questions, agreements and disagreements. They’ll then share those with a partner.

I’ll ask them to define “fulfillment” as partners and then as a whole class, and I’ll have a student record the various definitions we come up with on the board. We’ll see if we can find any common ground in those definitions before moving on, using a new color to circle the words and phrases whose implications seem to rhyme.

We’ll then read the next paragraph:

The common good…is the result of an “interdependence” that creates a network of social good that expands and has an impact on people. The common good is a “plus,” the result of interaction and mutual influence that connects various actions, initiatives, efforts and decisions. If we were to add up the individual goods, we could not explain the existence of this “plus” that transcends them and, at the same time, enriches them. (Paragraph 61)

I’ll ask students to take out their phones (a net negative in schools to be sure but very helpful at key moments in our program), open their TikTok For You page, and scroll through the first few videos. They’ll then tell their partners what those videos were and why they think the AI-powered algorithm chose those videos for them specifically.

I’ll ask: If we all watch the first video from everyone’s For You page up here on the doc cam, one by one, would that be the common good as the Pope defined it? Why or why not?

After some discussion, I’ll follow up: If we create a new, collective For You page that replaces everyone’s individual For You page, would that meet the Pope’s criteria? Is he arguing for a For Us page on TikTok?

We’ll then read from the next paragraph, the final excerpt for the class period:

Working together for the common good means having a shared vision. It is clear that there are many ideological and practical differences among people, as well as differing interests and frequent disagreements, but that does not mean it is impossible to engage in dialogue to establish a set of basic agreements that enable the creation of a shared vision, upon which everyone can move forward together. (Paragraph 62)

I’ll say: Imagine it’s the weekend and you’re meeting friends for lunch. What are the top options likely to be? I’ve seen a lot of Panda Express bags around the cafeteria recently — would that be a top choice? What else?

After students discuss in small groups, I’ll ask them to list the top three choices most likely to come up for them and their friends. We’ll then make a class top-three list on a new part of the board.

I’ll then say: The Pope isn’t saying the common good means everyone should order from their favorite spot and then meet for lunch together nearby. He’s arguing that people with different interests and frequent disagreements still need to create shared vision. How would that work here? How do we decide where to eat?

Students will discuss various methods of choosing. In my experience, this conversation often ends with students suggesting an online, randomized spinner or even coin tosses as a way to decide.

When they do, I’ll say: Randomness is tempting but it’s not allowed by the rules the Pope is establishing. We need to create a shared vision. That’s harder than flipping a coin, but as you get older you’ll run into more and more moments where coin flips won’t work. We need to start practicing now.

The Common AI Good

There are many challenges facing educators in the AI era: creating policies, designing strategies, exploring tools, maintaining analog moments, communicating with families, keeping up with the tech…the list goes on.

But the encyclical reminded me that there are some basic questions underlying many of those challenges, and that engaging with those subterranean issues would help us navigate the complex — and seemingly ever-shifting — landscape lying before us. There’s a reason the Pope didn’t get to the tech until Chapter 3.

One of those first-principle-style questions is how to reconcile the individual good and the common good. Students should be considering that question, and I’ve tried to show some possible ways to help them do so. But educators should be too.

As teenage-specific as some of those intellectual tripwires above are — How could the TikTok For You page be making it harder to pick a lunch spot with friends? — they are not too dissimilar from the situation in many schools right now. If every classroom, or even every school, has its own AI policy, what does that mean for the district? If some teachers follow the letter of the policy and others the spirit, what does that mean for students moving between those classrooms? Maybe most importantly, if every city or state is racing to be a model of effective and responsible AI-in-education, what does that mean for our country’s ability to make decisions about AI in schools?

I’d argue the current “techlash” — and the current parent resistance against AI and edtech in schools in particular — stems partly from a feeling that, to use the Pope’s language, we are failing to “engage in dialogue to establish a set of basic agreements that enable the creation of a shared vision.” The parents protesting and the AI labs proclaiming and the schools policy-making are not in conversation, and the notion of basic agreements hasn’t been broached in a serious and sustained way.

We have parallel monologues instead of dialogue. That’s confusing enough for all the adults involved, but what are young people supposed to make of their elders talking past each other yet again, as we did about social media a decade ago?

I certainly can’t claim to know how to make genuine dialogue happen between all those groups. But I do think one good place to start is where the Pope does, with the notion that there is a common AI good we can try to define together. And as teachers, we can start in our classrooms this fall, listening to our students create the For Us page we need.

Like many in the AI-in-education space, I find it hard to keep up with all the thoughtful content coming at me. I’m doing my best to read or listen to as much as I can, but two forces recently conspired to have me consider a parallel approach.

For one thing, it’s mid-May. Every teacher has a “fuel low” light flashing in their brain right now. I’ve also been hankering to push my sense of what AI makes possible, both to teach myself more about the tech and to better understand what it shouldn’t do even if it can.

So I recently worked with Claude Code (because that’s currently the only place you can use Opus 4.7 in “max” mode) to take on a mostly non-coding task. I gave it the names of some of my favorite podcasts, both about AI-in-ed and in general, and then asked it to search online for others like those. I then had it explain to me how to download the most recent two episodes of all those podcasts. While all this downloading was happening I was playing Candyland with my daughters before bedtime, which mostly meant brokering a deal between the older one and the younger one about how you can get unstuck from a Licorice Space.

I then had Claude take me through the process of transcribing all that audio. It showed me how to download and then run free transcription software. I asked it to read all the transcripts and suggest the ones I should upload to a new NotebookLM to create a meta-podcast. It pulled 10 to batch together, telling me NotebookLM’s podcast feature tended to lose the thread if it’s dealing with more than 10 sources. I took Claude’s advice while noting it seemed a little condescending about its competitor.

I uploaded those 10 transcripts to a new NotebookLM and then added some specific instructions in the podcast window before pushing “generate.” I then let NotebookLM whir in the cloud as I slept.

The next day I went downstairs to my apartment building’s gym to ride a stationary bike and listen. I often listen to podcasts as I ride, so on one level this time felt no different. I put on my earbuds, pushed play, and started to pedal.

The usual NotebookLM voices came on, one woman and one man, and they opened their usual simulated podcast banter. They then started digging through the ideas and details shared in the ten podcasts I had uploaded. They even argued with each other, with the woman claiming the superstructure of education was crumbling before our eyes and the man responding that this was a moment of evolution not collapse.

Soon though their synthetic binaries began to grate. I could almost hear the transformer architecture clicking away behind their words, like they were two figures revolving in and out of the door of a cuckoo clock, waving their arms mechanically at each other as they circled by.

I was getting to the hard part of the ride, so I needed something more compelling to listen to — real people talking. I switched on the latest episode of “Your Undivided Attention,” which featured Tristan Harris interviewing two guests with contrasting opinions about the cybersecurity and geopolitical implications of Anthropic’s Mythos model. Josephine Wolff, a professor of cybersecurity policy at Tufts University, was on one side. Fred Heiding, a research fellow at Harvard’s Kennedy School of Government, was on the other.

Harris went back and forth between the two, shaping their debate. Wolff spoke with an apartment door and sticky notes on a wall behind her; Heiding was framed by a bicycle on a wall and a large TV. Together the three of them explored the nuances of Mythos’s aftermath. This was much more compelling as both an intellectual experience and a workout distraction.

Then at one point, as Harris was processing his guests’ comments and considering his next question, he closed his eyes. It seemed like the kind of eye close I’ve been getting to know this semester.

As the AI conversation in general continues to reverberate, and the AI-in-education conversation gets louder and louder too, there is sometimes (often?) just too much to absorb. I find myself closing my eyes at certain moments in conversations with students and colleagues so I can pause the visual stimuli coming in and just focus my mind on the next thing I’m trying to say.

I may have been projecting, but I saw Harris having that experience on that podcast, as he tried to hold and articulate his guests’ evolving and opposing views. It felt like we both close our eyes to look at the same thing.

We’re currently being inundated by the general. LLM-produced or -informed content is a rising tide around us. These systems, having absorbed essentially the entirety of humanity’s visible thought, take in the specific context and prompts we give them and give us back a synthesis. They are, in a sense, gigantic versions of the meta-podcast I made for myself.

Being subsumed in the general will be a challenge for everyone, but it poses a particular risk for students. They need to engage with the specificities of other people’s minds so they can sharpen the specificities of their own. They need to find where they see eye to eye with others and where their perspective differs.

In my last few years teaching English, I brought Shakespeare’s King Lear into the curriculum. The motif of sight was a productive vein to mine. At one crucial moment, a son convinces his blinded, despairing father that they are standing on the edge of a cliff which is, in fact, not there. To do so, the son paints a vivid verbal image:

Come on, sir; here’s the place: stand still. How fearful
And dizzy ‘tis, to cast one’s eyes so low!
The crows and choughs that wing the midway air
Show scarce so gross as beetles…
The fishermen, that walk upon the beach,
Appear like mice…
…the murmuring surge,
That on the unnumber’d idle pebbles chafes,
Cannot be heard so high.

The son tells this lie to help his father see a truth, to help him live past his despair. But was that the right thing to do? Did it even work? Would they do that to their parents? Students would debate those questions vigorously in our class discussions.

One year I tried a rule that they could only offer their own opinion after they paraphrased the opinion of the person they were responding to. I hoped that they would learn what they thought by first trying to see their classmates’ point of view. I also knew how fast English class conversations can spiral into vagueness and unrelated personal opinions if they’re not grounded in the specific text at hand.

One point we could all agree on was that, in that scene, the blind father sees in his mind’s eye the world the son describes. He is inhabiting the world another person creates through carefully chosen words.

Here’s Claude’s summary of that scene:

A disguised son leads his blind father, who wants to die, to what he claims is a high cliff.

Here’s Claude’s gloss on it:

Sometimes loving someone means deceiving them.

Imagine students encountering either of those outputs instead of the speech itself. The intellectual work has already been done, and the human work is therefore no longer doable. Students would have ridden an electric bike to their destination and wondered how anyone could get out of breath. The particular is the friction that forces students to shift mental and emotional gears, which helps them realize that others do the same.

The first time I encountered that speech, as a high schooler, I decided that at the end Shakespeare was mimicking the sound of a wave hitting the shore:

the murmuring surge,
That on the unnumber’d idle pebbles chafes

“Surge” is one long syllable, followed by the multiple short syllables inside “unnumber’d idle pebbles,” followed by the next long syllable: “chafes.” I can still close my eyes and say those words out loud, seeing a wave come in, mix among the pebbles, and go back out.

I might just be seeing things. What matters is I was taught to look. I learned to engage with the words and ideas of others so as to better shape my own, which meant I saw the world as full of other particular people with particular perspectives. That insight matters even more now, as young people encounter words more often averaged than authored.

“Agency” is a recurring headline in both the AI conversation and the AI-in-education conversation.

For many, agents have replaced chatbots as the state-of-the-art in the AI world. Agents can be set off to complete tasks on their own, ideally without human intervention, and come back with the results. The theory, to state the obvious, is that AI is better off when it has agency.

Education has long held a similar theory about students. Recently, though, agency has taken on new urgency as those of us in schools try to ensure students can be actors, not spectators, in their careers to come.

I worry that the aspirational talk of “orchestrating agentic workflows” will in practice mostly mean people watching AIs doing things. That’s why I think it’s important to help students try to do things in real life, learning from both what they can and can’t make happen. Young people need to know what it feels like to succeed and to fail and to know that both are a kind of accomplishment. They need to know doing is something people do too.

The two students who organized the mayor’s visit showed me what that can look like this year. Watching them do things and fail to do things, and reflecting on how I had and hadn’t supported them effectively, helped me explore a pedagogy for agency.

Russell and Andrew (pseudonyms as always) came into senior year with something to prove, though they may not have realized that themselves until later. They wanted to show themselves and the people around them who they were becoming, not who they had been.

They started the year in the purpose-development and career-connected learning program I run exploring careers in law enforcement, first local police and then potentially FBI. Students test-drive careers in three-week loops in our program, and for the first loop the two young men just wanted to get in (even) better shape.

In their career research they had found that the Newark police academy has fitness requirements for enrollment, so they would often head into the hallway during my class to work on their one-minute pushups. One mistake I made during that time was to presume they were focused on the “one-minute” part of the pushup, but of course like many teenage boys (including me at that age) they were just worried about how many total push-ups they could do.

I only realized my mistake too late when the loop was over, but it became a reminder for me that authentic work towards career goals needs to actually be authentic. Otherwise students are exercising their agency in a kind of zero-gravity environment, with no real sense of whether they’re moving in the right direction.

Later, in our penultimate loop, Russell and Andrew had their first failure in the program. To start our second semester, they wanted to launch a “police academy” club in our school. For days, they used our class’s AI assistant to help them iterate on a club proposal and semester-long plan. I deliberately stayed mostly out of the way, so they could experience that process themselves.

They then pitched it to the club director at our school. He denied their request. His points were well taken; the students hadn’t made clear what the club would accomplish by the end, and it also wasn’t clear what made it a “police academy” and not a place to work out and watch police-related media.

They were discouraged. They had spent a few hours total working on the club idea. I was discouraged too, by the fact I had gotten the space-scaffolding ratio wrong.

The final loop was the mayor loop. Over the course of the year, Russell and Andrew had come to realize that police work happens within the context of city government and city policies. Russell also discovered he had a family connection to someone who works at city hall. (A good lesson in the power of personal networks.)

They leveraged our AI assistant again, this time to write the initial outreach email. They and I talked first through their goals for the email, then used the prompting and context skills we’ve been practicing to explain everything to the AI, then reviewed the AI’s output before sending it. They got a response, sent an AI-assisted reply they ran by me first, and got another response. A few days later they told me they had a date: March 3rd.

They then worked with AI to design the program for the mayor’s visit. It helped them draft questions about public safety and career exploration and helped them refine — but not write — their opening remarks.

The week before the mayor arrived, Russell and Andrew spent several class periods in a row rehearsing the entire visit with our network’s communications director. They told her their vision, and with her they practiced each part of it. AI had helped them make a map, but to learn the territory they had to walk through it.

The morning of March 3rd, seniors were abuzz. Mr. T, why are you wearing a suit? Russell and Andrew actually got the mayor to come? This is really happening today?

I assured them it was — I had pulled them all out of 4th period, 5th period, and lunch and ordered 17 pizzas for the occasion — but that I didn't know much about what would happen. Russell and Andrew were in charge.

The two young men greeted the mayor at the entrance and walked him to our meeting hall. I met him outside the hall. As he walked up with them I had my first moment of realizing the mayor was really here. I said, “Hello. Thank you for being here, Mr. Mayor. I’m the teacher of the class.” I stepped aside and gave Russell and Andrew space to walk by. They took it from there, going way past what any AI could do for them.

If you’d like to read more about what students asked and the mayor shared, here’s a local news story from later that week.

For me, the pedagogy story was that building student agency specific to the AI era means three things:

We need to help students experience authentic progress in an activity that has relevance for them, with success defined by those who specialize in that activity.

We need to balance space and scaffolding as students undertake long-term projects, so that they experience formative but not summative failure.

And we need to show students how to use AI tools to achieve personal goals while also showing them where the AI’s work stops and theirs begins — and how teachers and other mentors can walk that final mile alongside them.

When you put all those moves together, which takes as much trial and error by us as by students, you end up with the kind of reflection Andrew shared in that article:

“It made me feel like I’m capable of so much more,” he said. “People didn’t believe we could get the mayor to come speak in our school, and we did it.”

Emmanuel (a pseudonym as always) was giving his capstone presentation, which 11th and 12th graders in my purpose-development and career-connected learning program deliver throughout the spring. In these presentations, students share where they’re coming from, where they’re going, and how they’ll get there.

As part of that narrative, students share what they accomplished in the six “loops” they cycled through in our program. Here are some of the projects Emmanuel undertook this year:

• Loop 1: Creating a pitch for a student micro fund that would empower high school investors with $50 grants via philanthropic donations.

• Loop 4: Creating the proposal for a student-led course on day trading and pitching the proposal to decision makers at school. He used our class’s AI assistant to support this work since he’s never taught or created curriculum before.

• Loop 5: Launching and teaching that course, which met twice a week for all of second semester and was designed entirely by Emmanuel and his friend. About 10 fellow seniors signed up to take it.

• Loop 6: Finalizing the launch, with other friends, of an official 501(c)(3) nonprofit dedicated to supporting unhoused people in our local community. As they prepared to launch, they used Claude Code to create their website.

At the end of each presentation, the assembled audience (which usually includes the presenter’s friends from other classes, family members, and former teachers) shares affirmations with the presenter. Audience members raise their hands if they want, and get called on by the presenter, to answer this question: “How do you know the presenter is going to be successful?”

Several of Emmanuel’s peers shared how much he had matured during senior year, how his procrastination habits had gotten slowly better and how he was starting to find a path forward that he felt passionate about.

After audience members finish sharing their affirmations, I share final thoughts with the presenter. My goal in these moments is to help students see something about themselves they can’t see from the inside. Here’s the gist of what I said to Emmanuel (who also shared in his presentation that he loves basketball):

Let’s click back to that slide where Emmanuel shared all the projects he worked on each loop this year. Take a look at that again everyone, in case you missed it the first time. Look at those six projects. You too, Emmanuel. Look at that trajectory.

The NBA playoffs are happening right now. Emmanuel, I’m sure you’ve been watching right? Have you seen the Knicks’ recent run? They’ve been running up a historic margin of victory against their opponents. The Knicks are peaking at the right time. (Fingers crossed this post ages well here especially.)

You’re peaking at the right time too, Emmanuel. Earlier this year your late assignments outnumbered your on-time ones. It felt like you were adrift at times. But now look at that slide on the screen. You got a lot done in our program this year, and what’s most impressive to me is how you kept leveling up each loop.

You started off creating a deck pitching an idea that never got off the ground, and you’re ending the year having launched a nonprofit and taught a semester-long course. And watching you up here, speaking with confidence about these accomplishments, shows me this is just the beginning.

We all gave Emmanuel a final round of applause and he went out into the courtyard to take photos with the friends who had gathered to support him.

Yesterday I had my usual bout of FOMO regarding ChatGPT 5.5. The latest OpenAI model, which Ethan Mollick among others has touted as the new AI gold standard, was released a couple of weeks ago.

Like many others, I cancelled my subscription to ChatGPT after OpenAI’s decisions during the Pentagon-Anthropic showdown earlier this year. But GPT 5.5 has been singing a siren song since its release, leaving me barely hanging onto the mast.

As I’ve plumbed the depths of my FOMO, I realize that the intrigue for me is in seeing firsthand whatever new capacities and nuances the newest model provides. As soon as Claude Opus 4.8 comes out, I’ll feel relaxed again knowing I’m seeing the bleeding edge of AI unfold word by word on my screen. That is until Gemini 3.2 drops. Then GPT 5.6. And so on…

That AI-specific FOMO is commingled with the same feeling planned obsolescence has been conjuring for decades now. We discuss that phenomenon every year, and students are taken aback when they see the consumerist logic linking the 2027 Camry, NBA 2K27, and iPhone 17. They realize new models are a constant background in our lives, like the seasonal rhythms they attempt to emulate.

Those of us engaging in the AI-in-education conversation are particularly susceptible to FOMO of all kinds, because we want to ensure that we’re on the cutting edge and our students are too.

New AI models are developed according to “scaling laws,” whose proponents argue that the more data and compute companies put in, the more AI capability they’ll get out. Another way of putting it is that the more “experiences” an AI model has through the combination of training data and compute, the more it grows.

But those of us in education focusing on AI’s growth need to remember that the growth that matters, students’ growth, is already happening in our classrooms. AI can support that growth — Emmanuel used our class’s AI assistant on many of his projects this year — but only up to a point. Emmanuel needed to teach the class AI helped him design.

A different kind of scaling law applies in schools. Training data and compute become lived experience and reflection. Emmanuel went through six loops this year, each time trying out a different career path while remaining headed in the same general direction. And that was just in my classroom. He was having different ongoing academic experiences the other six periods a day.

Meanwhile he was having innumerable personal experiences, in the hallways in between those classes and especially in his life outside of school. He spoke in his presentation about how much he appreciates his relationship with his older brother, who he’s stayed connected to after his brother went off to college.

Even the presentation itself seemed like another ingredient in Emmanuel’s growth. He always gives me a handshake on the way out of class, and that day it felt a little different, like he was reintroducing himself to both of us.

On Monday morning, it was warmer to walk on the west side of the street in Newark. I was bringing five juniors in my career-exploration and purpose development class from our high school to a local courthouse to launch a new internship program. It was one of those chilly April mornings, so we walked on the side of the street where the sun was shining.

We spent about 90 minutes in the courthouse, getting a tour from the judge setting up this internship. I first got connected to this judge a few years back, when a student who was related to her helped set up a field trip to her courtroom. Since then we’ve made the field trip an annual occurrence and she and I have gotten to know each other. I’ve made sure to keep in touch in between our visits, something both my principal and my parents have modeled for me over the years.

So when the judge was looking to set up a new internship program at her courthouse, she reached out to me to see how we could collaborate. I told her I had just the students in mind, 11th graders who have been exploring legal careers all year.

A few weeks later, those juniors and I were greeted by the judge in the courthouse lobby. She took us around the building, introducing us to everyone from the jack-of-all-trades plumber to the public defenders to the payroll team to her fellow judges.

Every time we finished a conversation with someone, she repeated the same refrain to the students: “It’s important to know everyone you work with and treat everyone with the same respect.” After we finished talking with the plumber, she spent several minutes explaining how much he did around the building and how he was almost single-handedly making it a place befitting of the important work being done inside.

On our way back to school, the students and I chose the sunny side of the street again. One of them reminded me that I had almost brought my jacket for the walk until they convinced me I didn’t need it: “Aren’t you happier you’re not carrying it right now?” I was.

Soon after, I was suddenly delighted to realize we were going to pass a Starbucks on our walk back, so I happily put in a mobile order for some cold brew. (In K12 education, coffee is always calling, but its siren song might be most pronounced in late April and early May.) It didn’t feel right to get myself a coffee and nothing for the students, so I ordered them each what I get my daughters on Starbucks runs: one cake pop per person.

We stopped in to pick up everything and then finished the walk back to school chatting about what excites them most about the internship. The west side of the street had warmed up so much at that point that I found myself wishing for more shade.

Wednesday

Two days later, I was out for the day to help lead presentations on AI. In my absence, I had students start imagining their first year after high school via a custom bot. Here’s a demo conversation I ran to test it, based on a real student’s college and career interests this year (and mimicking the typically short AI inputs I usually see).

Are you a junior or senior?

junior

Nice — so you’ve got time to explore, which is a good thing. Two questions for you:

What kind of path are you leaning toward after graduation? College, trade program, military, workforce — even if it’s just a vague lean, that’s fine.

And what kind of career or work catches your eye right now? Doesn’t have to be locked in. Could be something you saw online, something a family member does, or just a topic that keeps pulling your attention.

college. I want to study law.

Law — solid. Let’s dig into that. Before we talk schools, I want to understand what pulls you toward law specifically.

Was there a moment, a person, a case, something you watched or read — anything that made you think that’s what I want to do? Tell me what sparked it.

We did a mock trial earlier this year in my class and I enjoyed the drama of it.

That’s real — the drama of it, the performance, the back-and-forth. That tells me something. Were you playing an attorney role in that mock trial, or a witness, or something else?

Attorney. I was defending someone unjustly accused.

So you weren’t just performing — you were fighting for someone. That combination is interesting: you like the theater of it and the purpose behind it.

Most colleges with strong pre-law tracks have mock trial teams that compete against other schools. That’s where you’d get that same energy weekly, not just once a year. When you’re looking at colleges, that’s one of the first things to search for.

Are you looking at any specific colleges yet, or still wide open?

I want to go to Howard.

Howard is a great fit to explore for this. It has a strong reputation for producing Black lawyers and leaders, and its proximity to D.C. puts you near federal courts, the Supreme Court, and major law firms.

What draws you to Howard specifically? Is it the culture, the location, the reputation — what made it stand out to you?

All of the above.

Here’s what I notice: you’re drawn to standing up for people and to being part of something bigger than yourself. Howard’s location in D.C. means there are likely opportunities to work with legal advocacy organizations or civil rights groups near campus, even as a freshman.

So we wouldn’t lose track of the ideas that came up in those conversations, I had students complete short reflections and follow-up ideas on a Google Form that same period.

Next lesson when we have a little extra time (amid the capstone presentations occurring most days now), I’m going to ask students to reach out to specific people associated with the organizations the AI chat steered them towards.

For example, I’d want to have that student above connect with someone at a local legal advocacy group in DC to start building relationships there should their future path end up taking them to Howard.

Then of course the key is for that hypothetical Howard student to actually travel to that office once they get on campus. Ideally, it will be easier for them to walk in the door and introduce themselves after doing a yearlong, weekly internship at a courthouse in their hometown. Plus they’ll walk in knowing to treat the receptionist and the CEO the same way.

The final block before we got back to school from the courthouse, the juniors and I hit a busy avenue. Even though they’re 16 and 17, I instinctively found myself walking a little more slowly as we approached the street, to make sure I was close to both the first and last student in the group. Then we crossed.

Every day for the rest of the year, each period is devoted to 1-2 student presentations. (For more context on the run-up to these, see my post from three weeks ago.) Students share where they’ve been, where they’re going after high school, how they’re going to get there, and who has supported them on their journey so far.

Students invite that personal community to join their presentations, so that each day our class seating chart is adjusted to accommodate visiting family, friends, and former teachers. Others often appear on FaceTime, and when students ask me to, I record the presentation so they can share it with those who can’t be there. After each presentation, students gather with their invited guests in a small courtyard outside my classroom and take pictures to commemorate the moment.

Today, for example, Shawn (a pseudonym as always) shared his long-term photography career plans with his stepfather and sister, Keyonna told her four best friends and father about her teaching plans, and Nathaniel explained how he was carrying forward his cousin’s legacy to his mother, grandmother, uncle, and friend since elementary school. They each brought their past into our classroom’s present to show us the future they’re building.

Before students share, though, they prepare. And those preparations have taken on new significance for me after a recent realization I had about where AI is heading.

Markdown files as digital dossiers

For a certain corner of the AI world, there’s a new imperative that echoes a much older one. “Physician, heal thyself,” has become “User, summarize thyself.”

I found myself in that neighborhood of the AI landscape recently, after Anthropic cut off my OpenClaw agents’ access to tokens and I turned to Claude instead to continue some work.

One of OpenClaw’s primary innovations was having an AI agent create a kind of record of its conversations with you on your own computer. These records, which are distillations of the most important and relevant information for your work, are stored in markdown (.md) files the AI agent leaves in the folder you’re working in. The clever idea here was that, by opening and rereading its own markdown file notes every time it talks to you, the AI is able to immediately get up to speed on your work (and becomes less susceptible to the memory lapses common in traditional AI conversations).

With my OpenClaw effectively decommissioned, I wanted to bring the project notes over to a new folder for Claude to reference as I picked up the work there. I realized I had several markdown files’ worth of project notes, but nothing about me.

I spent a few minutes talking into a voice note about what my goals are, how I like to work with AI, how I want AI to work with me, and a little bit about my personality and broader personal context. I then cut and pasted the transcript of that voice note into Claude Cowork and had it turn my musings into a markdown file on my laptop:

It occurred to me as I was doing this that I would probably want to use copies of that file in every project folder I end up opening, so Claude always knows who I am as we begin to talk.

It then occurred to me that this might become a de rigueur part of our AI lives moving forward, having essentially a personal dossier we bring with us to each new AI encounter.

Right on cue, I later saw that the popular state-of-AI podcast “The AI Daily Brief” had done an episode on just this topic a week beforehand:

Markdown self-portraits are going mini-mainstream.

Implications for education

This whole notion, creating text summaries of yourself and your user preferences to carry to each AI conversation, still feels more than a little surreal to me. And I know that even in AI-forward classrooms, including my own, it seems nigh-impossible to teach this concept by the end of this year when we’re still working on prompting and context engineering in standard chatbot conversations with prom fast approaching.

But I do think this is where the technology is going, which means if we want to make sure our students are ready for college and careers, we eventually need to make sure they know how to create and manage their own digital dossiers.

In other words, we need to teach students how to show up to AI. But they’ll be better at that, and better see that for what it is, if they first know how to show up for each other.

Presentation preparations

Students spend the last two weeks before spring break creating their presentation slides, and then they spend a day or two after break rehearsing before we start the presentations. Here’s how a recent rehearsal lesson went:

1. For a hook, we read excerpts from a recent NYT article about in-person skills fast becoming essential in the AI-era workplace. We then discussed how developments like this in the world of work make the skills we’d be practicing today more important than ever.

2. We watched an excerpt of this speech from a world champion of public speaking (1:32-5:15 to be precise). Before watching, I asked students to look for the techniques the speaker uses to make his message compelling. We then debriefed the public speaking moves the speaker made and recorded the list on the board.

3. I then put this slide up on the board and led students through 1:1 rehearsals following this protocol.

I closed the lesson by saying something to this effect:

Your stories are compelling. You’ve worked hard to make your slides compelling. The final step is to become compelling yourself. When your community gathers in this room to celebrate you and your future, you want to present not just your slides but also yourself and the person you’re becoming. Using these techniques is part of how you do that. And as we discussed at the start of class, being compelling in person is now a skill that matters even more in the careers you’re all exploring.

Next, the night before every student presents, they need to upload a voice recording of their final, full rehearsal to Google Classroom.

How we show up

Our students are graduating into a world in which they will likely need to distill themselves into markdown files to better use the AI tools waiting for them in their workplaces. They will need to become stenographers of themselves.

That will go better, though, if they first spend time deciding who they are and who they’re becoming. I started “Mike.md” by talking into a voice note to get across the essence I want AI to know. Having students do presentations like these is one way to help them start finding their essence.

And one key, I’m increasingly realizing, is making sure students take the moment of the presentation itself as seriously as the content of the presentation. The lived experience of presenting yourself powerfully to your people has a clarifying effect. And it’s a clarity, I hope, that students can take with them into their lives after high school.

The point here is not just that students will now need to transmute rites of passage into contexts for AI conversations, though I do think we need to help them — and ourselves — wrap our minds around that idea.

The more important point is that the AI moment increasingly requires us to ask an old question in a new time: “Who are you?”

We’ve always wanted to help students answer that question. And now, with AI podcasters espousing “personal context portfolios,” it’s even more incumbent upon us to make sure students know to start looking for their answers within themselves and their community. They need to know how to show up as themselves and for each other before they learn how to identify themselves to the machines.