Every edtech executive with >1,000 followers on LinkedIn (including yours truly) has published their #NotAllEdtech post arguing that we should focus less on screen time and more on screen value. These companies all argue we should focus on how kids use their products rather than how much.

Many of those executives should be careful what they wish for. Indeed, to observe how many products are used in classrooms would only validate the fears of many parents—screens that pacify rather than challenge; screens that isolate rather than connect; screens that decrease rather than increase a teacher’s visibility into a student’s learning.

Show Your Cards

I’d like each of my colleagues to show their cards. I’d like to see from everyone engaged in this discourse: one (1) stationary video of a full classroom session.

Speed it up if you want, but that video will do more to illuminate the real-world social and cognitive impact of these products than any company-funded research study, customer testimonial, or LinkedIn post. Just show us how the pieces fit together, how the humans, ideas, and technology add up to more than the sum of their parts.

Our Cards

Here is a video of NYCPS teacher Liz Clark-Garvey teaching an Amplify Desmos Math lesson called Sand Dollar Search. As a treat, I have coded each segment of the video for “what the students are paying attention to.”

How Our Education Technology Works

At Amplify, we know that, yes, certain enterprising students can learn quite well from an LLM or a library card. But most students benefit enormously from the motivation, accountability, and support they receive from their teachers. We also know that if you ask students “why do you put up with school?” the vast majority of them will say, “Because it’s where my friends are.”

So we use technology as a loom and weave together people and their ideas.

In Liz’s class, you’ll see students on their devices together. You’ll see them use those devices for short intervals—none longer than 8 minutes—13 minutes of screen time total. The devices first stir their thinking, letting them play with math in ways that are impossible with pencil and paper. Then the devices make that thinking visible to Liz who uses it in conversation with the whole class—calling kids to the board to elaborate their ideas, contrasting several ideas together, noting their similarities and differences, never speaking for longer than 90 seconds without checking in with students.

You’ll see students come to realize their work matters and react accordingly: working harder, participating actively, and learning more.

How Most Education Technology Works

Every edtech executive on LinkedIn seems willing to stuff at least one education technology into the wicker man and light it on fire. Everyone seems to agree that unrestricted access to YouTube is bad, for example. Everyone hopes this controlled burn will divert attention from their technology. Me, I hope the light from the fire helps everyone pay more attention.

With lots of education technology, students spend too much unaccountable time on their devices. Everyone works on different things. The dashboard gives teachers limited visibility into that work. Kids know that teachers can’t easily check up on them. They come to realize their work doesn’t matter and react accordingly: drifting off task, onto other tabs, and out of any state parents would recognize as “learning.”

I’m an edtech developer and a parent of elementary school-aged students and I welcome greater scrutiny of our industry. After winter comes spring—a time of growth and renewal. Many edtech companies will try to survive this winter by warming themselves next to a fire that is right now consuming several of their peers. But they should show their cards—show a stationary video of a single classroom—and let parents decide whether or not to use their products for kindling as well.

Parents value the human relationships that schools produce, relationships that support student learning and human flourishing. Everyone in edtech should show their cards. Are they weaving those relationships together or pulling them apart?

Los Angeles parents are fed up with schools loading up students with laptops and tablets, and assigning schoolwork on a slew of apps.

Some families, who had decided against giving their children screens at home, told school board members that they were appalled to find young students using school-issued devices — even in kindergarten. Some parents complained that their children were able to play video games or watch social media videos during school.

The edtech industry ignores at its own peril just how disenchanted students have become with their devices, especially after virtual schooling through the pandemic. Before the pandemic, it often felt novel and exciting to unload the laptop cart. Students got a little more dopamine and gave a little more attention simply because of screens.

Several years later, kids feel very differently. You can see the difference even when news outlets report on classroom use of artificial intelligence, surely the buzziest new education technology. Those articles invariably contain rapturous descriptions of personalization, dynamism, and the future. Yet you’ll see, without fail, a field photo of kids looking like they have been dosed with a veterinary-grade tranquilizer.

Edtech needs a new melody.

Teaching is the work of composition. Teachers compose the resources in a classroom—including education technology, but also paper, textbooks, questions, explanations, whiteboards, other students, the physical space, etc.—into experiences that help students learn. I find it helpful to think of those resources as musical instruments and their composition as a melody.

In many classrooms, students find the melody pretty dull—regardless of the teacher’s use of education technology. Few would argue this is a pleasing melody, for example. Every note is pure teacher explanation for a 20 minute segment of class.

Here is a different melody. Many teachers are enthusiastic about whiteboards and their potential for quick formative assessment. The teacher poses a brief question; students answer on whiteboards; the teacher responds to the class results with an explanation or summary followed by another question.

In that melody, the students hear several different notes—whiteboards, teacher questions, teacher explanations—all played to direct student attention towards learning.

Digital technology frequently produces as monotonous a melody as a teacher speaking endlessly in front of a blackboard. In the sort of experience reported in The New York Times above, you’ll often see the teacher explaining a task, then sending students off to work by themselves on their computers, often in silence, after which the teacher collects the laptops and ends the activity. The experience is digital, but the music is still a dirge. These devices are some of humanity’s most incredible achievements, but when students use them in schools they often hear only two notes—teacher explanation and screens.

We must learn other melodies for education technology. As one example, I taught a class last week here in Oakland, CA, using our curriculum and I can tell you that kids were on devices for no more than five minutes at a time before I brought them together to a) set up the next task, b) ask selected students to share their ideas, c) ask the class to help settle a dispute, d) do a closing assessment on paper, etc. The composition looked a lot like this.

In one of my favorite moments, students heard multiple notes simultaneously—a chord. Students were matching different sets of cards together to make equal products.

There were hundreds of digital cards around the classroom. With the push of a button, the software analyzed which card was the hardest for the class to group. I told the class, “This card is our hardest. Show your neighbor where you have placed it. See if you agree.” Students passed their laptops across the table to their classmate as an object of conversation just as you or I might a coffee table book.

This is a challenging way to teach, in part because computers incur heavy switching costs. It takes quite a lot longer to log into a laptop than it does to “log into” a paper handout or a class conversation. As we segued from computers to a class conversation, I begged the kids to only close their laptops halfway, because if they closed it all the way, the district laptops automatically logged them out and we’d lose significant time logging back in. Those costs all mean education technology needs to bring more value to a learning experience than any analog substitute. In my case, I used a digital card sort because it could find out the most challenging card with speed and accuracy that I can’t match on my best day with paper cards.

“How much screen time is too much?” is an important question, but just one of many. We should also wonder what students do with their time on screens and what purpose that work serves. With so much edtech today, the instruments are shiny and new but the melody is still monotonous. We should wonder how we can play a more interesting melody with computers in class, kids switching in and out of them as easily as switching between people in a conversation, between pages in a book, between instruments in a symphony called learning.

I spent the morning at a middle school recently, watching three sixth-grade teachers each teach the same math lesson from Amplify Desmos Math. Those teachers and their work helped me understand that curriculum has a current and you’re either swimming with it or against it.

First, Here’s The Lesson

In the activity I watched—Order in the Class—every student received a card with a number on it.

We didn’t choose these numbers randomly. Rather, we’re trying to invite and develop early incorrect ideas students frequently develop about numbers. For example, the idea that 9.45 is greater than 9.6 because it has more digits following the decimal.

Students had to move around the room and find other students with a number greater than theirs, with a sign opposite theirs, etc, and at the end arrange themselves in a single line from least to greatest. That last part was pretty spectacular to watch.

Why I Think This Lesson Worked

In my head, I played this lesson against a worksheet that had all the same numbers, with kids sitting and deciding if their number was greater or less than a bunch of other numbers. Certainly, kids who understood the math at a certain level could rip through that worksheet faster than the social experience.

The social experience seemed so much more effective, however, for a few reasons.

1. The paper can’t check you like your classmates can. Every new conversation about numbers is an opportunity to negotiate your ideas with others. Over time, you’re likely to find a person who disagrees with you. That’s a valuable check for understanding.

2. Negotiating your ideas deepens your understanding. The student who can rip through the worksheet should have to explain why 9.6 is greater than 9.45 and in doing so strengthening their own understanding.

3. Students aren’t suffering for worksheets in math class. They are suffering for opportunities to move and talk.

How The Teachers Swam With the Current

Every curriculum—every product and tool teachers use—has a current, an expression of how its creators think things are and ought to be. Sometimes that current moves quickly, reflecting very opinionated creators. Other times, when its creators are more agnostic, the current moves slowly.

I watched each of those three teachers find the current in our curriculum and start swimming with it. One of those currents is the conviction that students should become the curriculum, that at different points in a lesson student ideas should become objects of study for the class. I was happy, then, to see:

• One teacher had a very effective way of saying, “I’m totally confused here,” and drawing out the class’s ideas in order to help unconfuse him.

• Another teacher dealt herself a number card and played with the class.

• The third teacher heard an idea from a student, one that was correct but imprecisely expressed, and stuck with it, studying the kid as the kid said, “So when the number gets bigger but the sign is negative the number is actually smaller.”

There are sometimes good reasons to swim against the current in your curriculum. Maybe the current fundamentally doesn’t suit you. You can always kick across or against it, but you won’t go as fast or as far as swimming with it. If I had to offer new teachers any advice in swimming with Amplify Desmos Math it would be this:

Sensitize yourself to the moments when students reveal their curriculum. A kid will say something that might seem out of bounds at first but which is actually working to stretch the boundaries of what we call math. In those moments, stay present. Express surprise, delight, or interest, even if it feels unnatural at first. Swimming feels unnatural at first. Ask the kid to say their thought one more time. Tell the class to tune in. Tell the class that something very interesting is happening that you want to understand better. Ask a second kid to restate the first kid’s thought even if just to buy yourself time to think. Ask the class to use their hands to signal whether they agree or have questions. You do not need to stay in this moment forever—just long enough to convince one more kid that they too might be the curriculum and to ready yourself for the moment when they share it.

This is how you find the current in our curriculum.

PS. On Tutoring

I have been working in the same eighth grade class every Monday all school year, generally with a group of boys who I can’t say enough good things about. They are energetic, outgoing, and kind to one another in ways that defy my expectations of eighth-grade boys. They also frequently need help with eighth-grade math, help which the state of California certifies that I can provide!

You might think, as I did, that this is a fantastic arrangement. But I have not found it easy at all to make that help available to these boys. Part of that difficulty results from the facts that this class speaks Spanish and all of those boys speak Spanish better than me, facts which I suspect have dimmed their impression of me. But another part of that difficulty is that the relational work of tutoring is just grinding, a mix of pushing (“¡okay ándale!”), prodding (“dígame más de este número aquí”), and encouragement (“¡estos estudiantes aquí ay!”).

Anyway, last week, the kid who kind of runs the group asked me what my name was. “What do I call you if I want your help?” he said.

“Mr. Dan,” I said.

There are four weeks left in the school year and I finally have—for now! tentatively!—enough trust and goodwill to help them with math.

I mention all of this as an invitation to anyone who feels excited about AI performing this tutoring work to help me understand the world as you do. What experiences have you had tutoring children that lead you to believe this is possible?

Upcoming Presentations

Let’s hang out this summer! At each of these events, I’ll be describing how to make math a more creative discipline for more students, and how to support teachers in that work.

• Florida Association of Mathematics Supervisors. June 23, 2026. Jacksonville, FL.

• Florida Council of Teachers of Mathematics Annual Conference. June 24, 2026. Jacksonville, FL.

• Math in the Sun. July 1, 2026. Scottsdale, AZ.

Odds & Ends

¶ I saw this somewhere on Twitter. What’s your move here if you see this answer? I’ll post the teacher’s response in the comments.

¶ I had a very nice chat with Craig Barton about maths and AI. He and I grew up as teachers at about the same time. We both turned social media into professional development. I think Craig and I are both generalists, preferring to make connections across teaching, math, and technology broadly without specializing maximally in any. I might have Craig wrong there, but for all of those connections, it was one of the most interesting conversations I’ve had about math edtech this year and, as a bonus, Craig recorded it. Here are his five takeaways from the conversation.

¶ A new YouTube channel from Amplify offers some awesome lil PD bites. Like, rate, and subscribe.

¶ Wisdom from Phil Hill:

If AI anxiety is what finally opens serious conversations about learning quality, institutional purpose, and the thirty-year drift toward transaction, that is worth something regardless of how the technology itself plays out.

¶ Great summary from Jill Barshay of research indicating that AI gives students different feedback depending on how the researcher described their gender and race.

The researchers found consistent patterns across all the AI models. Essays attributed to Black students received more praise and encouragement, sometimes emphasizing leadership or power. (“Your personal story is powerful! Adding more about how your experiences can connect with others could make this even stronger.”) Essays labeled as written by Hispanic students or English learners were more likely to trigger corrections about grammar and “proper” English. When the student was identified as white, the feedback more often focused on argument structure, evidence and clarity — the kinds of comments that can push writers to strengthen their ideas.

¶ Houston ISD is turning nine schools into AI-focused schools. What could that mean? “Minimal details have been released by the district as to the day-to-day instruction at Future 2 campuses.” Meanwhile, New York City put plans for a similar school on hold after parent outcry.

Pivoting edtech towards humanity means using the power of technology to align one human’s desire to learn with another human’s desire to teach.

For Example

The physical classroom.

A bunch of people want to learn. A bunch of people want to teach. Until we connect them in time and space, those people are misaligned, their desire wasted. The technology of the physical classroom brings those desires into alignment.

Digital media.

As a new teacher, I noticed kids had a desire to learn how math connected to the world outside the classroom. I had a desire to teach them about those connections but very little ability to do so because the world outside of the classroom was outside and we were inside. The technology of digital projectors and cameraphones let me capture the world outside the classroom and bring it inside for our analysis. Technology brought our desires to learn and teach into alignment.

For Counterexample

If you ever have the feeling that edtech isn’t all that interested in humanity, it’s frequently because:

Edtech companies try to align human desire to the power of technology.

Edtech companies often take a particular technology as the answer and then retrofit teaching and learning into the question. This is why, for years, various companies insisted that teaching is something very close to playing a video of an explanation, which makes “just play YouTube videos” seem like the answer to the question “why is teaching hard?”

Edtech companies ignore second-order effects.

A student feels like their class is moving a little slower than they’d like. An edtech company then suggests having every kid learn on computers which let them work at different paces. The company ignores the second-order effect that “kids also like learning together and now they can’t.”

Misalignments That Interest Me Currently

Kids want to work on paper and it’s hard for teachers to know what they’re doing.

Teachers struggle to support a student’s thinking if it isn’t visible. Lots of thinking happens on paper and teachers often lack the time necessary to review and respond to it. How can we make that paper-based learning more legible for more teachers?

Coaches want to support teachers and teachers want their support.

Coaches often have too many teachers on their roster to support with model lessons and walkthroughs. Also, many teachers want support but not in the form of a model lesson or walkthrough. The desire to give and receive support are misaligned here.

Teachers want support in leading whole-class discussions.

Whole-class discussion is some of the most satisfying work for teachers and productive learning for students. But it is very hard work. Misaligned.

Caregivers want to support their kids but don’t know how.

Many parents and caregivers want to do more to support their kids’ learning than they do currently. But they often lack visibility into student learning and may need some education themselves. The reports that schools send home are frequently summative, low resolution, and a waste of ink or pixels overall. Teacher emails are much more useful but time-consuming for the teacher. What can schools and edtech companies do to help align caregivers, teachers, and kids here?

Who will do this work?

I could point to dozens of people doing this work of pivoting edtech towards humanity. They are exceptional. Many of them are my coworkers. Common to each of them is an excitement for new technologies and a desire to understand the work of teachers and the lives of students that I can only describe as “insatiable.” If that’s you, let me know what you’re working on in the comments.

Featured Comments

My obituary for Khanmigo and AI tutors inspired so many of you to share your own stories of grief. This newsletter is here for you. This was a common interesting interaction. From deep within their grieving process, someone would ask:

What’s the alternative? If Khanmigo doesn’t work, what’s the Plan B?

I’d point out the effect of Saga’s tutoring interventions in Chicago Public Schools as one of many interventions that has had a positive effect on student learning. Still grieving, this person would respond that this intervention is “difficult to implement at scale.”

This is such a strange standard for evaluating interventions in education. It is very true that good things are often difficult and expensive while useless things are often easy and cheap. Many people mistake this fact as an argument for doing useless things! (My colleague Chris Blackett develops this idea more at Talent Lab.)

Mae Baltz mentions another kind of subsidy Khanmigo frequently received: administrative mandate.

As a teacher in one of the areas that received money for Khanmigo I was asked to have my students interact with Khanmigo at least 10 times per month (each student).

In spite of those mandates, Khan Academy reported yesterday that “only around 15% of students who have access to Khanmigo engage with it.” That indicates pretty serious misalignment.

Katelynn Petersen describes the difference between AI and human tutors. Please write this down somewhere!

As soon as I started hearing about tutors being replaced by AI, I knew that the people responsible for such nonsense had never tutored a day in their life. 40% is remembering to ask about the novel they are writing, the tea they spilled about their friends, or the language test they’ve been studying for all year. 40% of my time is spent just building confidence and reassuring students they’re doing the right thing. 20% is actually teaching math.

Odds & Ends

¶ I have very little to say about Khan Academy’s new venture, announced just before I posted my obituary last week. An edtech visionary is stymied by traditional education and retreats to the friendlier terrain of corporate e-learning. Am I talking about the $10,000 degree Khan Academy will offer in partnership with ETS and TED? Or am I talking about the $7,000 degree Udacity offered in partnership with Georgia Tech after their disastrous experience trying to support college freshmen 13 years ago. Answer: yes. Hop in the time machine, kids. NB: Read Glenda Morgan’s pre-mortem or John Warner’s polemic.

¶ Marc Watkins writes about the same crisis of purpose in higher ed that I am seeing in a local eighth-grade class.

It’s easy to dismiss lazy students or burned-out teachers turning to AI, as many seem to do in the comment sections of social media posts, where we hear a litany of solutions from folks that range from bluebooks to oral exams to entire technology bans. But AI isn’t simply a crisis in assessment. No, the true crisis here is purpose.

¶ A couple of tremendous writers and thinkers take on the AI chatbot tutor’s promise of “infinite patience.” First, John Warner talks about his gratitude for the finite patience of his teachers.

Some of my most important formative educational experiences involved some teacher or authority figure losing patience with me.

Second, Julia Freeland Fisher asks, in a world of infinite patience, “whose while are you worth”:

AI’s champions often laud it as “infinitely patient.” AI’s unerring support is undoubtedly powerful, especially when time and resources are scarce. But it falls short of the experience that accompanies real patience: not just material support, but the feeling you are worth someone else’s while.

¶ I’ve worked in curriculum development for over a decade and this comment from Stanford’s Sam Wineburg on Justin Reich’s podcast earlier this year is a really rare insight.

Ultimately, curricula are not for kids. Curricula are for the teachers. And if the teachers don’t feel exuberant and don’t feel ennobled by being the mediators and the adapters of those curricula, they can be the best and most thought-out curricula in the world, but they’re ultimately going to find dust on some shelf.

¶ I’ll have more to say about the digital backlash someday. For now, I’ll let it suffice to say three things.

1. I think the coalitions that are forming are among the wackiest I’ve ever seen on any issue.

2. As a parent, I wish schools would scrutinize their use of edtech more closely.

3. I think the edtech companies that understand teaching and learning, that prioritize the humanity of teachers and learners, are probably going to be fine.

Khanmigo, Khan Academy’s AI chatbot tutor, died last week. Because Khanmigo is software, it does not die in the traditional, organic sense. You’ll still find it in the sidebar of Khan Academy practice exercises. But Khanmigo as an idea, as a shorthand for the edtech industry dreams of software that tutors as well as humans, has died, crushed underneath the expectations of its own creator, Sal Khan.

“For a lot of students, it was a non-event,” Khan told Matt Barnum last week in Chalkbeat, referring to Khanmigo’s release three years ago. “They just didn’t use it much,” he continued. Khan Academy’s Chief Academic Officer, Kristen DiCerbo, gave a similar assessment: “So far I am not seeing the revolution in education.” These assessments could not be more different from the expectations Khan set out for Khanmigo at its birth.

Sal Khan brought Khanmigo to life from the TED stage three years ago, just as he did mastery grading 11 years prior and video lectures four years before that, each time predicting an imminent revolution in education. Khan claimed that Khanmigo, and AI tools like it, represented “probably the biggest positive transformation that education has ever seen.”

Khan’s announcement of Khanmigo generated millions of views and precipitated widespread media attention, a book deal, and a steady flow of philanthropic and government subsidies. Newark Public Schools received $25,000 from the Bill & Melinda Gates Foundation for Khanmigo licenses. Indiana schools could apply for $50,000 in state funds for Khanmigo use. Palm Beach County received a $2,000,000 commitment from the Stiles-Nicholson Foundation to help schools purchase Khanmigo licenses. Microsoft eventually sponsored Khanmigo licenses for every teacher in 49 countries.

These subsidies breathed life into Khanmigo, but they also raised difficult questions. Could Khanmigo breathe on its own? Would schools and districts purchase Khanmigo at an unsubsidized price? If Khan’s predicted transformation were at all imminent, with students learning new ideas at faster rates and teachers experiencing previously unknown levels of support, why wouldn’t Khanmigo sell itself?

Curiously, Khanmigo began to assert itself more aggressively in the Khan Academy student experience. Over its short, three-year life, Khanmigo grew from an unassuming circular avatar in the lower-right corner of Khan Academy, to a circular avatar that would expand and, just one second after pageload, ask the student if they needed help.

In 2026, Khanmigo became an always-on chatbot experience, already activated with or without the student’s invitation. A Khan Academy spokesperson told Barnum they made this change because “students were not seeking out Khanmigo’s help as much as we had hoped.” Some began to wonder why, if its value were so obvious and transformational, did Khanmigo have to assert itself ever more aggressively into the student experience?

Even as it announced itself more loudly to students and teachers, Khanmigo began to perform farther and farther below the expectations of its creator. Sal Khan predicted that by the end of 2024, AI tools like Khanmigo would cut “90% of teachers’ admin tasks.” One year later, after that prediction had clearly not come to pass, he would extend his prediction an extra ten years to 2034. Khan Academy’s user projections also declined dramatically—one month predicting “a million or two million” Khanmigo users in a year and the next month predicting half that with “500,000 to one million students.”

On a recent webinar, Kristen DiCerbo indicated that student usage of Khanmigo was not what Khan Academy wanted. “I will tell you, we see more ‘IDK IDK,’” she said, “more passive kinds of interaction than we would like.” Critics suggested that the difference between Khanmigo and human tutors were vast, with the chatbot unable to draw on a relationship with students, unable to initiate or end conversations with the sensitivity possessed by even quite average human tutors.

At this point, with Khanmigo already struggling to meet Sal Khan’s expectations, critics began to scrutinize Khan Academy’s efficacy research. Laurence Holt dubbed its deficiencies “The 5 Percent Problem,” so called because Khan Academy’s strongest effect sizes (0.26 standard deviations above the mean in a 2022 study, for example) were achieved only after excluding 95% of the study population. In a more recent study, Khan Academy lowered their inclusion threshold significantly, avoiding the 5 Percent Problem, but also watching their effect size boil away to nearly nothing.

Khanmigo was raised with a silver spoon in its mouth, possessing advantages completely unknown to most other edtech chatbot startups. It had some of the earliest access to OpenAI’s generative AI technology. It had the backing of a hyperscalar in Microsoft and abundant cloud computing credits. It had the endorsement of national, state, and local officials. It had the phone number of some of the wealthiest people in the world. If Khanmigo died in spite of those advantages, what hope then should the rest of the edtech industry place in chatbot tutors?

In the past, Sal Khan has blamed teachers for low student Khanmigo usage, saying that teachers “need to figure out ways to engage them more” with Khanmigo. In Chalkbeat, DiCerbo blamed students, saying, “Students aren’t great at asking questions well,” which will come as a surprise to anyone who has ever known a small child. Yes, it’s possible that some of the most inquisitive beings on earth aren’t all that great at asking questions, but it seems more likely that chatbots like Khanmigo aren’t all that great at inviting, understanding, or answering those questions.

These are stages of grief, and Khan, himself, seems to have moved towards “acceptance.” He now says, “I think our biggest lever is really investing in the human systems,” with technology playing a supporting rather than leading role. This is doubtlessly a more fruitful path for education technology, as edtech historians like Larry Cuban and Justin Reich and edtech critics like Audrey Watters have argued for decades. It remains to be seen, however, if this path will appeal to Khan’s benefactors in the technology industry. Will they be as excited to support human systems as they have been software that tries to abstract humans away from human systems?

Indeed, given that Sal Khan has tried unsuccessfully for nearly two decades to abstract humans away from human systems—first with human explanation, then with human evaluation, and most recently with human tutoring—it seems unlikely that he is the right person now to pivot edtech towards humanity. Instead, it seems more likely that he should sit the next decade out and spend that time learning everything he can about the humans at the heart of the system that, for two decades, he has tried and failed to transform.

2026 Apr 25. Here’s what I think it looks like to pivot edtech towards humanity.

They know they can complete the next slide or not. They know they can answer the next workbook question or not. They know how to hide. Whether they’re working on screens or paper, they know how to look busy and when.

I’m watching the teacher circulate around the room. As she moves, every student inside a two-desk radius locks in on their page, brows furrowed, scribbling or typing, but when the teacher moves on they relax.

This is the challenge—making difficult cognitive work matter to kids. Teachers have lots of responses to that challenge, but most of them scale the time cost linearly with the number of students. Checking in with students in class. Grading each warmup. Grading each student for participation. Calling home. Grading all the essays. Double the students and you double the time cost for each of those interventions. This is burnout. Teachers need non-linear responses here, responses where doubling the students costs something far less than double the time.

“I have been struggling with this post-COVID,” veteran educator Nick Corley told me recently. “The number of students who only work when watched or for a grade has increased significantly. I’d love to hear what is working for others.”

Five Friends

Me too, Nick. So I pulled in five friends—veteran educators, coaches, total pros, and asked them:

What is a reliable way you can let a bunch of students—students who outnumber you 20 or 30 to 1—know that their effort, their work matters?

Zak Champagne taught elementary math for years and recently became Chief Content Officer at Flynn Education. He said the key to getting kids to sustain their effort with students was giving kids permission to stop sustaining their effort.

Now this might sound counterintuitive, right? Walking away from a math problem sounds like they don’t have to show effort. However, it was my experience that when the young people in my room knew they could walk away, they were less likely to do that. And they would spend more time working through the tough stuff. Sometimes just knowing you have the option to walk away from something is the very thing that keeps you going. And providing that option communicates to them that you trust them.

Katrine Bryan has been a secondary math teacher in San Luis Obispo County, CA, a math teacher coach, and I’ve showcased her teaching here before:

Moving away from a viewpoint focused on teacher knowledge to a stance that celebrates student voice and student actions will build the foundation, showcasing that what a student does matters: to the teacher, to their peers and ultimately, to themselves.

Pam Seda has held down jobs from classroom teacher to district math supervisor. She consults and recently co-authored the book Choosing to See: A Framework for Equity in the Math Classroom, which seemed relevant to my question.

Students learn that their work matters when the audience shifts from the teacher to their peers. When a classroom moves from a group of individuals to a community of learners — where anyone can be an expert and anyone can contribute to the group’s collective knowledge — the stakes become real.

One example: when students do group work, instead of asking each group to present their own process to the class, have the class speculate about how a group was thinking based solely on the work they see. No presentation, no narration — just the work, doing the talking. In this way, the phrase “make your thinking visible” becomes far more meaningful when students know their peers — not just their teacher — are the ones doing the reading.

Dylan Kane is one of the most thoughtful and discerning math educators around, one who teaches at a high level and still somehow has energy to share his thoughts at his newsletter:

I think really hard about the first things students do when we’re launching into a new activity to build momentum. I often start with a simple, straightforward question on relevant prior knowledge, and ask students to answer on mini whiteboards. Students are most likely to see their learning matters when I meet a few basic conditions: I build off of what they already know, I begin with something students can do successfully and that makes them feel smart, and I check to make sure students are learning and adjust when they are not. Mini whiteboards are the best tool I’ve found to help me meet students where they are, adjust on the fly, and build momentum at the start of class.

Tracy Zager is a former classroom teacher, current coach, author of Becoming the Math Teacher You Wish You’d Had, and perhaps most importantly, the editor of my book (still forthcoming):

I need to do the slow and careful and authentic work of creating a community where each student’s thinking and work really does matter–to the student, to their classmates, and to the learning of the class as a whole. My first thought is there are three big components:

• I need to choose curriculum and tasks that are worthy of my students’ time and thinking.

• I need to give a why behind everything we do–what is the purpose? If I don’t have a good purpose for a task, into the bin it goes.

• And I need to teach students how to listen and learn together.

Three of my five friends name the same solution—community development—the slow, halting process of investing one student into the presence and ideas of another. This strikes me as one of those non-linear responses. As the class size grows larger, so too does the size of the community, the number of people who are, potentially, vested in your work. It’s also a solution that runs counter to the prevailing wisdom that what these kids all need is their own individual AI tutor.

I’m trying not to be a reactionary here, complaining about Kids These Days and yelling at clouds, but there does seem to be a dramatic vibe shift post-COVID that I am sure is multi-faceted and I’m grateful to these five friends (and anyone in the comments) each working to understand that shift and work to change it.

Odds & Ends

¶ Tim Daly recently brought useful data to the question of what it takes to get kids to try hard.

What makes kids try harder? Teachers, mostly. Strong teachers motivate students to elevate their effort as the material gets more challenging. A positive school culture - the sum of many teachers and support staff aligned to the same standard - ensures consistency across classrooms and magnifies the effect.

¶ EdSurge reports on a perspective towards AI among teachers that I also find quite common—not optimism, not pessimism, just indifference.

When teachers consider introducing AI tools to students during class time, the calculations they make change. The relevant question becomes: What student learning problem does this tool solve? Many educators are still trying to answer this question, even after several years of exposure to generative AI in some capacity.

¶ Bill McCallum, a lead author of the Common Core State Standards and the Illustrative Mathematics curriculum, is publishing a newsletter. In a recent post, he re-opens a 20-year-old paper from Sweller, et al, celebrating direct instruction and criticizing discovery learning. McCallum checks the citations and finds something interesting in Sweller’s praise for worked examples:

Here is the irony. The strongest evidence-based use of worked examples—carefully designed, presented in contrasting pairs, with structured opportunities for analysis and discussion—looks a lot like the kind of instruction that Kirschner et al would dismiss as constructivist-based minimal guidance. It manages cognitive load, yes, but through thoughtful task design, not by eliminating the need for student reasoning. It is, in fact, a form of productive struggle.

¶ Just a bonkers survey out from RAND. Student use of AI for homework help is up 14%. The majority of those students use AI in spite of their belief that it’s hurting them:

As of December 2025, 67 percent of students endorsed the statement “The more students use AI for their schoolwork, the more it will harm their critical thinking skills” — up more than 10 percentage points from ten months earlier.

¶ Peps Mccrea writes about the ways lesson plan design and user interface design inform one another.

The principles behind great UI are often relevant to the classroom. We could even think of teaching as Learner Interface Design. Here’s what that looks like:

Teachers have to think ten steps ahead of 30 other people, imagining dozens of possible futures, many of them quite bad. It’s fantastic preparation for the work of designing technology.

OpenAI has done something that is, once again, very neat, but not useful for novice learners:

Research suggests⁠ that visual, interaction based learning can lead to stronger conceptual understanding than traditional instruction for many students. When learners can manipulate variables and instantly see the effects, they may be better able to internalize the relationships behind mathematical and scientific concepts.

Type particular questions about math and science into ChatGPT and you’ll see its usual chat, but also an interactive visual display where students can drag sliders back and forth and watch a diagram change. The experience feels frictionless, which maybe sounds great if you work in software, but this is a liability for novices, not an asset.

The OpenAI website contains examples from math and physics. Drag the sliders around yourself. The experience is silky smooth. You can make diagrams morph and dance almost without thinking about them, which is exactly the problem.

Novices need more friction, not less. Learning results from friction. It is a grinding of gears. You learn when you make your old knowledge and new knowledge lock together, reconciling what you know with what you knew. And at every point in that process it is easy to tell yourself, “Yes, I have done it. I have internalized the relationships behind mathematical and scientific concepts,” even if you haven’t.

When a novice uses OpenAI’s Pythagorean Theorem simulation, for example, they are much less likely to say, “Ah, I now understand that the sum of the squares of the legs of the right triangle is equal to the square of its hypotenuse” and much more likely to say, “Ah, when I move my hand right and left the square gets bigger and smaller.” Without friction, the superficial insight replaces something substantial. The empty carb replaces the whole grain.

Education research offers us some insight here. Lots of studies demonstrate that immediate feedback is more helpful than delayed feedback, particularly when we’re talking about giving the same test results immediately or several days later. But OpenAI’s diagrams are different. They have more in common with the early LOGO microworlds, where you’d make a turtle move around at your mathematical command. Simmons and Cope (1993) found that, in those worlds, students were more likely to use procedural strategies like trial-and-error when they received immediate feedback than delayed feedback. Kids needed friction to cross the threshold from superficial to deep understanding.

That finding is why, in our curriculum, we offer students a dynamic interactive and then follow it with something quite static, a question that asks students to slow down, to go beyond their easy conclusions about the interactive, a question that introduces more friction not less. When answering those questions, novices benefit enormously from an agent telling them, “This is good, but let’s think harder about what’s going on.” And, in 2026, there is no agent more effective at coaxing students to endure and engage in that friction than a human teacher. AI can assist those teachers in that work, certainly, but it is still humans who help humans do hard things.

With these interactive diagrams, OpenAI has yet again offered the world a great gift, though perhaps not the one they intended: an opportunity to better understand the challenges of learning, the sophistication of teaching, and the value of human relationship. These interactive diagrams are something that AI can easily do, but they are not, on their own, something that students actually need.

Featured Comments

Timothy Burke on AI saving teachers time:

But also because there isn’t any evidence of the repurposing of the time saved. That is the yawning void in the discourse of AI boosters: exactly what are they imagining that the saved time will be redirected towards that is a more valuable activity for trained professionals? The reason they don’t want to concretize it is because even if AI automates certain tasks effectively (a big if) AI boosters don’t actually know enough about any workflow or work processes in any existing profession to envision what professionals would rather be doing nor do they understand anything about what the obstacles to doing that work actually are.

So much of the tech and business community believes teaching time is fungible like money. Save money on your car insurance and spend it literally anywhere else. Save time grading papers after school and ... you can’t actually spend that time in school. If AI saves teacher time, well that’s great on its own, but too many people assume that time can transfer from outside to inside the classroom.

Upcoming Presentation

San Diego County friends: on Friday, March 27th I will be at the San Diego Educator Symposium and Publisher Fair hosted by Amplify. Space is limited so register by Friday, March 20, if you’d like to attend.

Odds & Ends

¶ Wess Trabelsi, a lament about education today:

[Generative AI] is crashing into our classrooms at the exact moment when traditional assessment proxies are broken, and more importantly, when students’ hopes about the future are incredibly brittle. We are asking students to willingly engage with the “desirable difficulties” of learning, in the hope that they develop the skills required to direct and audit AI output. Except, we can’t promise them a stable job or a predictable future in return, and they know it.

¶ Freddie deBoer, along similar lines as Trabelsi:

All of this education [in crisis] discourse (all of it, all of it, all of it) is downstream of the reality of the neoliberal turn, globalization, and deindustrialization. We decided that we didn’t want jobs that don’t require a college degree anymore, many people are not academically equipped to get a college degree, and so we manufactured this “crisis.”

We need to name clearly what teachers can and can’t do, and then to wonder why different groups of people tell teachers, “You can fix poverty.” Freddie deBoer makes a compelling argument that a) this isn’t something education has ever done anywhere on a national level, b) there are plenty of other levers—redistributive economic policy and social welfare, for example—that have.

¶ Get in the car, folks. It’s time for a rebrand!

Precision learning is fundamentally different [from personalized learning]. It would enable educators to use technology, data and evidence to identify exactly where a student is struggling, which interventions are most likely to work and how to deliver them effectively and equitably.

Call the program whatever you want. My questions will always be the same: a) what math does it let kids do? b) how does it make use of human relationship? With personalized learning, those answers have been a) math that reduces to a number or multiple choice response, b) before and after the program but not during.

¶ Kip Glazer is a helpful voice—the principal of a large school in the tech center of the world; forward-thinking with AI; realistic about the work of teaching, learning, and leading. She has been putting edtech companies on notice this whole month:

The best tool in the world fails when the workflow around it is broken. When it does not talk to your SIS. When it adds three steps to a process that already had too many. When it was designed without accounting for the fact that a teacher’s day is not a controlled environment. It is a living, unpredictable, deeply human one. [..] Human variability is not an edge case. It is the entire job.

¶ 🎉 S/o to my colleague Ana Torres for winning Best Female Hosted Podcast from the American Writing Awards. Beyond My Years is a great time.

By the end of 2028, what percentage of non-interpersonal teacher activities (lesson planning, grading, and parent communication) will teachers routinely delegate to AI systems?

My own prediction here would have landed somewhere quite a bit south of 50%, mostly because I think the interpersonal and non-interpersonal tasks of teaching are pretty tough to disentangle. Wess Trabelsi won with a prediction of 65%. (Congrats, Wess. I liked your entry and described it to the other judges as “a wild ride.”)

Nearly every entrant, including Wess, relied on a survey from the Walton Family Foundation and Gallup that surveyed teachers on their AI use and time saved. The big finding of the report:

Teachers who use AI weekly save 5.9 hours per week — the equivalent of six weeks per school year. Currently, about three in 10 teachers are using AI at least weekly, with more frequent users experiencing greater time savings.

I want to offer a few reasons why I suspect that finding is wrong. In fact, if it is anywhere close to right, if this technology (which I regard as “neat”) can effectively shave off six weeks of teaching work (which I regard as some of the most taxing that exists) then I have drastically misunderstood AI or teaching or both.

Here is why I’m skeptical of the finding.

People aren’t great at self-reporting the time they save with AI.

The Gallup data is entirely self-reported. Teachers were asked how often they used AI and how much time they saved on different teaching tasks. Meanwhile, the AI research firm METR went beyond self-reports with a similar study of software developers. They randomized a set of tasks between “AI allowed” and “AI disallowed” groups. They asked for self-reports of completion time, but they also measured actual completion time.

The AI-using engineers believed their tasks had taken them 20% less time than the non AI-using engineers when in reality it had cost them 19% more time. So I’m happy the AI-using teachers feel like AI has shaved off six weeks of their work, but the METR study should make us question self-reported data of this sort.

These sample sizes are quite small.

The most common response to the question, “How often do you use AI tools to do this teaching task?” was “never” for every teaching task surveyed. Only 4% of the 2,232 surveyed teachers described using AI weekly to “analyze patterns in students’ learning,” for example. That amounts to 90 teachers whose responses were averaged across other categories and then “multiplied by the number of contracted weeks per year (37.4, on average)” to get the time savings expressed as weeks. “Margins of error for subgroups are higher,” says the methods section, which I believe. Unfortunately, those margins were then multiplied by 37.4.

Okay also: saving time is not per se good!

Students seem to be saving tons of time these days using AI. Yet plenty of adults are more worried than excited about those savings, worried that students are saving time not doing work that they should do—not writing their early drafts of papers, not struggling to remember which solution method is most appropriate, not committing knowledge to memory.

We should worry similarly about teaching. Are we happy or sad that teachers are outsourcing their tasks to AI? Which tasks? It’s true that teachers can save time giving feedback on student writing by asking an AI to do it. If saving time were our only prerogative, we could save even more time by simply not assigning essays at all. Clearly, the more stuff AI can do, the more we need to wrestle with the question, “What stuff should AI do? Are there costs besides time that we should consider?”

Featured Comments

Michelle Kerr on last week’s teaching dilemma:

The tricky part isn’t explaining [whether the answer is “6-n” or “n-6”]. It’s getting them to stop and think about it before they answer. This is also the case with whether the slope is positive or negative or whether (x+3) has a zero at 3 or -3. It’s always about getting them to stop and take a beat before they answer instead of just mentally flipping a coin. The hard part is reminding them that they have the knowledge already if they just take a second to stop and tap into it.

Ryan Muller believes the MOOC comparison doesn’t account for AI’s vastly different technological power:

The scaling that has unrelentingly continued to extend its capability on ambiguous and long running cognitive tasks hits a brick wall against... teaching?

Just so I’m not misunderstood: yep, that’s exactly what I think. Humans have—to date!—needed other humans to help them do hard things they’d otherwise rather not do. So you can file teaching alongside therapists and gym trainers as professions where the impact of AI is going to be limited compared to the rest of the economy.

Upcoming Presentation

• March 13. This month I’ll be giving a keynote address at California’s annual math conference in Bakersfield, CA. We’ll be talking about why and how we can restore creativity to math, a discipline frequently perceived as a creative wasteland.

Odds & Ends

¶ Congrats to Amplify’s math curriculum team for earning an all-green rating from EdReports for Amplify Desmos Math, a curriculum that is near and dear to my heart.

¶ You simply have to hand it to Pam Burdman sometimes. Perhaps you recall the thunderclouds above UC San Diego’s math results recently. The Wall Street Journal’s editorial board called it “A Math Horror Show” attributing increased enrollment in remedial math to “increased admissions of students from ‘under-resourced high schools.’” (UC San Diego made the same attribution.) But Pam used the Wayback machine and found another factor contributing to that increased enrollment, one which I haven’t seen reported anywhere else.

¶ Rick Hess interviewed me for Education Week. One Q&A:

RH: What’s the one big thing a school or system leader should know when it comes to AI and schooling?

DM: The introduction of generative AI has not changed the fundamentals of your work: getting absentee students back to school; making sure kids feel supported and known; communicating results and challenges with parents; creating a positive working environment for teachers; and keeping kids safe. For all of its power in the world outside of schools, generative AI has not transformed the reality of any of those challenges and may, in the case of student mental health, exacerbate them. The work is still the work.

¶ I love the lines John Warner quoted from How College Works.

The fundamental problem of higher education is no longer the availability of content, but rather the availability of motivation.

Human contact, especially face to face, seems to have an unusual influence on what students choose to do, on the directions their careers take, and on their experience of college. It has leverage, producing positive results far beyond the effort put into it.

If human contact is a lever, we can think of technology as one kind of fulcrum. We see different results for kids depending on where we put that fulcrum and what kind of force we apply to the lever.

¶ Michael Pershan is correct: “most people do not want this.” Downthread he writes:

Most kids are not sitting around wishing that they could just hunker down with a tutor. They would rather be in class with their friends.

You can try to convince people to want something they do not want. You can yell into the wind. You can swim against the river. But the most prominent boosters of personalized learning do not seem to engage meaningfully with the nature of the wind or the river, with the reality that most people are not like them, that most people do not want this.

¶ Related: Ben Kornell ran an interesting webinar on the state of AI tutors. Notably, Khan Academy’s Kirsten DiCerbo said, “I will tell you, we see more ‘IDK IDK,’ more passive kinds of interaction than we would like.” That checks out. Andrea Pasinetti, CEO of Kira, offered a useful summary of the “micro-interactions that happen between students and tutors that can pull students out of the illusion of tutoring.”

Any latency. A question of two to three seconds delay between when a student asks a question and when a student gets a high-quality answer can cause distraction, it can cause a student to move on, it can cause a student to lose their own context in terms of what they’re learning and why they’re asking a particular question. An incorrect answer. A duplicative answer. An answer that maybe is not leveled properly in terms of reading level or concept difficulty or an answer that presumes knowledge that the student doesn’t have can all have a very negative impact on engagement and ultimately learning outcomes.

¶ I gave a talk at UC Davis last week and loved seeing the Better Call Saul-style advertisements for legal advice for (alleged but let’s be real) AI plagiarists on the bulletin boards.

This Week’s Question

Here is a 30-second video of a teacher from the 1999 TIMSS study.

Teacher: Okay what if I say, “six less than a number.” Six less than a number. Michelle?

Michelle: 6-n

Teacher: “6-n.” What do you think, Aubrey?

Aubrey: n-6

Teacher: Why do you think that?

Aubrey: Um because …

Teacher: You’re right. Tell me why.

Aubrey: Six less than a number.

Teacher: Right, do you see the difference?

In that clip, a teacher is trying to help students understand how to convert English sentences into algebraic expressions. It really seems to me the teacher has found herself in a miserable kind of jam—one that I think is recognizable to any math teacher with >0 days of teaching experience.

What My Friends Said

First, I asked my friends to describe that jam

Shelley Carranza is a high school math teacher in Mountain View, CA, a former math teacher coach, and a former colleague of mine at Desmos and Amplify:

The dilemma is that we don’t know whether Aubrey and her classmates really understand why the answer is n-6 instead of 6-n. Aubrey looks doubtful at the end of the video, and now I’m curious to know how many students are in the same place as Aubrey, wondering whether they’ve got the order right.

Jenna Laib is a math coach in Brookline, MA, and has developed the idea of a “Slow Reveal Graph”:

The teacher seems to anticipate a potential misconception: that students may recognize “6 less than a number” as subtraction but follow word order when creating an expression, producing 6 - n instead of n - 6. The first student called on did exactly this. Rather than engaging with the response, the teacher seemed to invalidate it and ignore it, moving to another student who provided the correct expression.

Fawn Nguyen is another colleague at Amplify. She helps people imagine a transformative math program at their school and has also been a math teacher and teacher coach.

The dilemma: The teacher was listening for the correct answer in a way that’s “n minus six or death.” And I say this with full empathy—English is tricky, Dan. Tricky for Michelle, and hard for a perennial English learner like me too. I wasn’t even sure which expression was correct until the teacher confirmed Aubrey’s answer and I heard that it was simply the reverse order of what Michelle had said.

That’s it. Two common teacher imperatives are in tension here.

1. The teacher wants to know what kids know.

2. The teacher wants the thing kids know to be the right answer.

Those imperatives have created this jam where the teacher finds out that a kid knows a wrong answer and then moves onto another kid hoping to find the right answer, doing (I suspect) some damage to the first kid’s idea of math and of themselves as a mathematician. How can the teacher get out of this jam?

Shelley Carranza:

At this moment, I really want to write both expressions on the board, and celebrate what the students know about the problem. From there, I’d want to give students a chance to discuss how you could decide which was right, and make sure to elicit the strategy of testing specific numbers.

Marilyn Burns is a former teacher, an expert in K-12 math learning, and an author of (I’m estimating here) 1,000 books about learning math:

To write an expression that represents 6 less than a number, some students think it could be “6 – n” and others think it could be “n – 6.” Then, for both options: Turn and talk with your neighbor and then we’ll talk about it as a class.

Stephanie Blair has held every job there is in K-12 schools except (I think) cafeteria worker. She worked with me at Amplify and Desmos, and now supports schools as they adopt Snorkl:

Instead of asking what the answer is, give students 2–4 possible correct answers and then have them decide and defend which one is correct.

Jenna Laib:

Here are two ways this could have gone differently:

(1) Debate: elicit multiple responses from students. Accept them neutrally, and record them on the board to support discussion. The format encourages students to justify their thinking.

(2) Try it out: stick with the initial response of 6 - n, and test it with a number. What is 6 less than 10? Is 6 - 10 the same thing? Record everything on the board.

In both cases, the goal is to make student thinking visible and support justification of why an expression works.

As an editorial aside, this problem may have been avoided entirely if the students had been using mini-whiteboards, because then this particular teacher would not have called on the student with the incorrect response. However, I’d rather encourage rigorous engagement with all student ideas!

Fawn Nuguyen

Write both expressions on the board. Ask students to think quietly first: which one matches “six less than a number”? Then turn and talk to a neighbor. Then rate your confidence —100% or nah? Now convince me.

Your Turn

Exercise for you, the reader, who I also consider a friend:

What is common among all of my friends’ suggestions—both pedagogically and socially?

Each of my friends have identified a common pedagogical technique but they also share a certain understanding of the social relationship between teachers and students. They have different imperatives. Great stuff. Thanks, friends.

Featured Comment

Efrat Furst on my review last week of the Stanford AI+Education Summit:

I keep coming back to the MOOCs story, I just can’t figure out how people refuse to see how similar it is and learn the lessons. It was just 10 years ago, we were all here to witness the rise and fall.

Audrey Watters identifies the stakes here:

There will be no “AI” tutor revolution just as there was no MOOC revolution just as there was no personalized learning revolution just as there was no computer-assisted instruction revolution just as there was no teaching machine revolution. If there is a tsunami, it’s not technological as much as ideological, as the values of Silicon Valley -- techno-libertarianism, accelerationism -- are hard at work in undermining democratic institutions, including school.

Odds & Ends

¶ Alpha School, the $65k private school that claims to have replaced teachers with AI, had a no good, very bad week. 404 Media interviewed former employees and reviewed documents and found that Alpha School:

• used AI to develop some sloppy, hallucinatory instructional materials,

• generated those materials, in part, by scraping content from other curriculum providers (including my own company FWIW),

• created clones of other edtech platforms like Khan Academy,

• exposed webcam videos of students at public URLs.

Check my post on LinkedIn for a bit more commentary but, putting it plainly: Alpha School is speedrunning some of the worst excesses of the move-fast-and-break-things era. Even still, I think we should separate a few questions:

1. Is Alpha School pursuing their model of schooling in a sloppy, unethical way?

2. Who does this model of schooling best serve?

3. Is there anything the rest of us can learn from it?

#1 is, barring some kind of contrition from Alpha School, a settled question.

Michael Pershan wrote a piece about the second question that managed to get agreement from everyone—critics and proponents alike.

This is a school that believes that the “core” of schooling should be taken care of as quickly and painlessly as possible so that the rest of the day can be opened up to things that actually matter. Most schools don’t do this! We instead tell kids that history is a way of understanding ourselves and others. Math, we say, can be an absolute joy, full of logical surprises. We tell kids that a good story can open up your heart and mind. Alpha doesn’t.

Dylan Kane wrote a piece about the third question, arguing that, whatever we can learn from Alpha School, it isn’t anything about technology.

¶ Congrats to fellow Desmos and Amplify alum Christopher Danielson for winning his third Mathical Book Award. I have gifted How Did You Count and its beautiful photos of everyday mathematical collections to a bunch of my friends when they become parents.

¶ Amplify colleague Shira Helft’s last statement of belief as a math teacher is cryptic and essential: “If you can, use a knife.” Read what she means.

¶ What happens when an AI bear hangs out with the AI bulls? Listen to my recent chat with Ben Kornell of the Edtech Insiders podcast.

¶ The party is sobering up. The triumphalism of 2023 is out. The edtech rapture is no longer just one more model release away. Instead, from the first slide of the Summit above, panelists frequently argued that any learning gains from AI will be contingent on local implementation and just as likely to result in learning losses, such as those in the second column of the slide.

¶ Stanford’s Guilherme Lichand presented one of those learning losses with his team’s paper, “GenAI Can Harm Learning Despite Guardrails: Evidence from Middle-School Creativity.” His study replicated previous findings that kids do better on certain tasks with AI assistance in the near-term—creative tasks, in his case—and worse later when the tool is taken away. “Already pretty bad news,” Lichand said. But when he gave the students a transfer task, the students who had AI and had it taken away saw negative transfer. “Four-fold,” said Lichand. What’s happening here? Lichand:

It’s not just persistence. It’s a little bit about how you don’t have as much fun doing it, but most importantly, you start thinking that AI is more creative than you. And the negative effects are concentrated on those kids who really think that AI became more creative than them.

A paper I’ll be interested in reading. This was using a custom AI model, as well, one with guardrails to prevent the LLM from solving the tasks for students, the same kind of “tutor modes” we’ve seen from Google, Anthropic, OpenAI, Khan Academy, etc.

¶ Teacher Michael Taubman had the line that brought down the house.

In the last year or so, it’s really started to feel like we have 45 minutes together and the together part is what’s really mattering now. We can have screens involved. We can use AI. We should sometimes. But that is a human space. The classroom is taking on an almost sacred dimension for me now. It’s people gathering together to be young and human together, and grow up together, and learn to argue in a very complicated country together, and I think that is increasingly a space that education should be exploring in addition to pedagogy and content.

¶ Venture capitalist Miriam Rivera urged us to consider the nexus of technology and eugenics that originated in the Silicon Valley:

I have a lot of optimism and a lot of fear of where AI can take us as a society. Silicon Valley has had a long history of really anti-social kinds of movements including in the earliest days of the semi-conductor, a real belief that there are just different classes of humans and some of them are better than others. I can see that happening with some of the technology champions in AI.

Rivera kept bringing it, asking the crowd to consider whether or not they understand the world they are trying to change:

But my sense is there is such a bifurcation in our country about how people know each other. I used to say that church was the most segregated hour in America. I just think that we’ve just gotten more hours segregated in America. And that people often are only interacting with people in their same class, race, level of education. Sometimes I’ve had a party one time, and I thought, my God, everybody here has a master’s degree at least. That’s just not the real world.

And I am fortunate in that because of my life history, that’s not the only world that I inhabit. But I think for many of us and our students here, that is the world that they primarily inhabit, and they have very little exposure to the real world and to the real needs of a lot of Americans, the majority of whom are in financial situations that don’t allow them to have a $400 emergency, like their car breaks down. That can really push them over the edge.

Related: Michael Taubman’s comments above!

¶ Former Stanford President John Hennessy closed the day with a debate between various education and technology luminaries. His opening question was a good one:

How many people remember the MOOC revolution that was going to completely change K-12 education? Why is this time really different? What fundamentally about the technology could be transformative?

This was an important question, especially given the fact that many of the same people at the same university on the same stage had championed the MOOC movement ten years ago. Answers from the panelists:

Stanford professor Susanna Loeb:

I think the ability to generate is one thing. We didn’t have that before.

Rebecca Winthrop, author of The Disengaged Teen:

Schools did not invite this technology into their classroom like MOOCs. It showed up.

Neerav Kingsland, Strategic Initiatives at Anthropic:

This might be the most powerful technology humanity has ever created and so we should at least have some assumption and curiosity that that would have a big impact on education—both the opportunities and risks.

Shantanu Sinha, Google for Education, former COO of Khan Academy:

I’d actually disagree with the premise of the question that education technology hasn’t had a transformative impact over the last 10 years.

Sinha related an anecdote about a girl from Afghanistan who was able to further her schooling thanks to the availability of MOOC-style videos, which is an inspiring story, of course, but quite a different definition of “transformation” than “there will be only 10 universities in the world” or “a free, world‑class education for anyone, anywhere” or Hennessy’s own prediction (unmentioned by anyone) that “there is a tsunami coming” for higher education.

After Sinha described the creation of LearnLM at Google, a version of their Gemini LLM that won’t give students the answer even if asked, Rebecca Winthrop said, “What kid is gonna pick the learn one and not the give-me-the-answer one?”

Susanna Loeb responded to all this chatbot chatter by saying:

I do think we have to overcome the idea that education is just like feeding information at the right level to students. Because that is just one important part of what we do, but not the main thing.

Later, Kingsland gave a charge to edtech professionals:

The technology is, I think, about there, but we don’t yet have the product right. And so what would be amazing, I think, and transformative from AI is, if in a couple of years we had a AI tutor that worked with most kids most of the time, most subjects, that we had it well-researched, and that it didn’t degrade on mental health or disempowerment or all these issues we’ve talked on.

Look—this is more or less how the same crowd talked about MOOCs ten years ago. Copy and paste. And AI tutors will fall short of the same bar for the same reason MOOCs did: it’s humans who help humans do hard things. Ever thus. And so many of these technologies—by accident or design—fit a bell jar around the student. They put the kid into an airtight container with the technology inside and every other human outside. That’s all you need to know about their odds of success.

It’ll be another set of panelists in another ten years scratching their heads over the failure of chatbot tutors to transform K-12 education, each panelist now promising the audience that AR / VR / wearables / neural implants / et cetera will be different this time. It simply will.

Many of our debates about the way math teachers should teach are what we might call “referred” debates—a debate whose true origin lies elsewhere. Debates that present as pedagogical—e.g., Direct or inquiry-based instruction? Team BTC or TLAC?—often mask debates about the nature of mathematics and the role of a teacher.

Consider, for example, these edits of two lessons from two teachers.

TIMSS Teacher

In this edited video of a teacher from the 1995 TIMSS Video Study, you’ll hear him frequently ask questions like, “Are you sure about that?” in response to a student’s incorrect answer.

Liz Clark-Garvey

Then consider this video of Liz Clark-Garvey teaching Grade 7 Amplify Desmos Math in New York City.

In that video, you’ll hear Liz say things like:

How would we use Tzameer’s method for Rectangle B?

I could be wrong but I think Saki is using Lizzie’s method.

I bet Gael is doing more of the Joely strategy.

You might think Lizzie’s method is more efficient. You can debate that. That’s fine.

How does each math teacher define “math” and “teacher”?

What good is a debate about their teaching techniques before confronting the different ways they both seem to conceive of the discipline of “math” and the role of a “teacher?”

Is teaching the work of instruction and evaluation? Is it the work of research and development? Are teachers also students? Is math a set of canonical, deductive ideas, each one falling naturally from the last like rain drops to the ground? Is math a set of dispositions and habits? An ability to recognize, name, and use patterns? A conviction that math must make a mark on me but I must also make a mark on math?

Perhaps “math” and “teachers” are all of those things, but I suspect you nodded more strongly at some of those questions than others. Our debates about “how” would lose a lot of their heat, I think, if we first defined the “what” and “why” of teaching math. It hasn’t done us much good, so far, pretending we all define “math” or “teacher” the same way.

Odds & Ends

¶ Which shape weighs the most in each sketch? No numbers. No operations. Lots of math. Thanks to Michael Pershan for reminding me of this problem set from blog-era all-timer Paul Salomon.

¶ Here are some slides and handouts from Anand Bernard’s talk at Utah’s state math teacher conference last week. Who is Anand Bernard? No one I knew before last week, but I attended his talk and I liked it a lot. He and I nominally covered similar ground at UCTM—how to help students learn and love learning math—but Anand covered it on the level of classroom routines and rituals, which has never been a particular area of strength of mine.

¶ The DebateMath Podcast took on the question, “Is AI Helpful in Math Education?” Great guests and conversation. I’m still thinking about this line from Cal Armstrong (who was “pro” on the motion overall):

I can barely get the students to read the entire problem and now they want me to make them read the responses of a chatbot that never shuts up.

¶ The math education world needs more long-form writing, so I’m glad to see my Amplify colleague Shira Helft sharing hers. One of her non-negotiables about math teaching:

There is a lot of heated debate about what is important to teach and know in math, and why it matters in the first place. I don’t always know where I fall, but I do believe this: if the math that students are doing doesn’t make sense to them, something is wrong.

¶ AWu tried to understand teacher use of AI by researching online teacher communities and interviewing teachers. Their analysis is no less valuable for being anecdotal. I thought their description of “The Formatting Tax” was especially useful:

Every time a teacher reformats AI output, they are reminded that this tool doesn’t know my classroom. It doesn’t understand my context and it produces generic content, expecting me to do the translation.

¶ Khan Academy, Stanford University, and the University of Toronto just published a study of 200,000 students, their use of Khan Academy, and its effect on their learning. I think Sal Khan’s description is accurate:

The study finds a statistically significant link between Khan Academy use and student learning gains, even after accounting for student motivation, teacher effects, and what was happening in the classroom year to year.

And I also think Justin Reich is correct to poke at the practical significance here:

If 20 years and $100MM+ in research and development builds an online math practice problem system that supports typical learning gains of 0.03SD, what can we expect from other edtech products? What can we expect from AI improvements to edtech products?

[..]

It’s not nothing! Let’s celebrate the incremental gains where we find them! But man, we’ve been working on this online practice problem model for 40 years, and I’m not sure the ROI is panning out here. If we think AI can help, what’s your bet on the value add? If you think AI can make something like KA 2x better, which is probably impossible, then you are still measuring effects with microscopes. You really need to be getting 5x or 10x gains to see viability, which-just as with magical ponies-is very fun to imagine.

While I have a lot of criticism of Sal Khan’s broader vision for math, teachers, and students, I do think he and his team have taken the “online math practice” genre as far as anybody. And all of that energy and effort has amounted to a blip for the average student. What’s next?

We took a different approach and I wanted to share a few notes about how my team did and didn’t work on this feature here at Amplify, because I’m proud of us.

Start with the problem not the solution.

Early 2023 was a time of profound foolishness in edtech. So many people started with the Large Language Model as the solution and worked backwards to define a problem. These people would hold a microphone to their face, making their voice louder on purpose, and then say things like:

• Hallucinations are good, actually. Fact-checking a computer is a game that kids and teachers have always wanted to play.

• It’s good for kids to talk to chatbots impersonating historical and literary figures. This is an ideal way to teach history and literature.

• It’s good for kids to ask chatbots to help them with a first draft of their essay. Ideally, we would not ask students to do this work of checking their understanding and organizing their thinking.

This was all untrue, of course. Experts in learning and history and writing were emphatic that these uses of an LLM would inhibit knowledge formation, reproduce biases, and undermine critical thought. But when all you have is a Large Language Model, you start to see the problems of learning as Large Language Model-shaped.

Meanwhile, I spent early 2023 basically the same way I spend every year: trying to understand the problems of teaching and learning, talking to teachers, substitute teaching monthly, working in classroom weekly, binging classroom video, and interviewing our front-line teacher coaches. This was slow work, but it helped me better understand a real, rather than imagined, problem of teaching and learning: facilitating whole-group discussions.

Make a solution by hand.

I started playing with the idea that teachers had too much information about their students, that all of that information was a blessing but also a burden. What kind of resource would help teachers make good use of that information?

At this point, lots of companies were willing to give teachers ten pages of analysis about their students, applying clustering methods from Natural Language Processing, drowning teachers in slightly less information but still drowning them, leaving them without an obvious “So what?”

I plucked ten classes out of our database at random, classes that had just finished one of our assessments. Then I looked at student thinking in those classes and asked myself, “If I were this teacher’s assistant, how would I help them make the best use of five minutes tomorrow?” I tried to assume this teacher was operating on maximum stress and minimum extra time.

Then I prototyped the first Discussion Moment in Google Slides.

You can see the similarities. Several student responses. A question to frame them. A narrative on the side. All made with love and by hand. Then I emailed it out to those random teachers.

I just made this slideshow for you based on your class’s work on the 8.5 end unit assessment. I’m curious if you think it’d be useful for you. If you have any thoughts I’d love to know.

I would have considered a single negative response or a bunch of non-responses enough reason to scrap the prototype. Instead, the responses from these strangers were pretty effusive.

I appreciate you sharing this with me. I love it!!! I really appreciate how it summed up the responses in my class and then gave me follow up questions for them. Can I share these with my class as I would love to have discussions based around the prompts given?

Yes, yes you may!

All of that investigatory work is slow. But I needed to know: is this thing even the right shape? Is this something teachers even want to hold in their hands? Does it fit their grip? Are we asking them to do something unnatural? Are we taking them off their “desire path”? Are we working for or against their aspirations for their work? That work can feel quite slow, until you consider the cost of making the wrong thing.

Now try to automate & systematize.

We asked ourselves, “What could we do if we only had seconds of turnaround time rather than hours? Which parts of that manual process could we pass off to a machine and at what cost to quality?”

• Constructing the discussion slide? Advantage: computers. A computer can populate a slide faster than a human and without loss of quality.

• Deciding how to construct the discussion slide? Advantage: humans. Computers could make some fine guesses here, but this work is largely a matter of taste, and we hire people with great taste. Asking humans to do that work has required a non-trivial but one-time investment across a countable number of activities in our curriculum.

• Picking responses from a class that match a given search criteria? Advantage: computers. We tested LLMs here and found their accuracy quite high, provided they were taken quite close to the target by a human.

We turned this division of labor into a set of authoring interfaces for humans and a set of API calls for an LLM.

Measure big & small & qual & quant.

We’re running a pilot with 200 teachers. That includes a pre- and post-survey. Our user experience team will conduct interviews afterwards. And we also spent a lot of time setting up our digital eventing and logging system, so during the pilot we’re able to answer questions like:

• Is this particular Discussion Moment activating as often as we’d want?

• Are teachers opening it?

• Were the right responses selected?

And then respond to those answers daily with product changes.

Patience is product.

It is so much easier to do this slow work when you have patient company leadership. Several months after ChatGPT made its splash, our competitors rushed to market with the exact same set of features. Chatbots. Autograders. Practice set generators. Lesson shorteners. Etc. Not bad ideas, but definitely the same ideas, not one of which, to my knowledge, has ever cost us a sale. (Seriously, I could count on zero hands the number of teachers who have asked me why we don’t have a lesson shortener.)

In spite of the massive pressure from the tech sector to appear forward-thinking, to do something, our board and leadership gave our product team plenty of time to make our thing, to work slowly and carefully, to make a marriage of technology and teaching that is uniquely Amplify’s.

Mailbag

Dylan Kane is currently on a No Tech January which makes his kind feedback on our tech a little extra special:

I’m impressed with the feature. I was skeptical when Dan alluded to working on something with LLMs a while back, but this is a solid use case and I think will add some solid value for users of the Desmos curriculum. Great work from the Desmos team.

Classroom discussions are very important and very hard. Recently at Amplify, I led the development of a feature that will make them much easier. Over the next couple of posts, I’ll describe the what, why, and how of “Discussion Moments.”

Every discussion teaches kids about math—and about themselves.

Among many other reasons, discussions are important because they’re moments when the teacher assigns value to students. In a discussion, the teacher says, “Hey—I have precious little time to teach what I know. Still, I’m going to dedicate some of that time for you to share and talk about what you know.” That’s a moment when students learn about math, but also that their own ideas have value.

Discussions are difficult, and “more wait time” is rarely the reason.

There are a few reasons why discussions frequently fail, and it’s rarely because the teacher didn’t give students enough “wait time” to respond, as is commonly believed.

1. The question was hard to understand or find your way into. For a long time, I’d ask my kids at dinner, “How was your day? What happened?” And my kids wouldn’t have much to say. Lately, I ask them to tell me two things about their day that happened and one thing that didn’t, and we all guess which was which. It’s an easier prompt, one that kids can find their way into with ease and then use as a launching pad into a larger conversation.

2. There isn’t enough to talk about. If your math class consists of a lot of binary, right/wrong questions, what is there for anyone to talk about? “A lot of us got this one wrong. Here’s a pie chart that shows how wrong we were. How about I show you how to do it?” That’s fine, but it isn’t a discussion, and it’s quite often a very dreary classroom environment for children.

In Amplify Desmos Math, a curriculum I work on, kids generally have plenty to talk about. Our interactives stir a kid’s imagination for even the most abstract areas of math. For example, this submarine interactive stirs up a kid’s ideas about adding positive and negative integers.

And then we ask kids, “Hey, what do you think about the star at +5? Can you come up with something that none of your classmates do?”

Let me tell you: Kids accept that challenge.

3. There is too much to talk about.

This is a good problem to have, but it’s still a problem. In the class screenshotted below, 25 students have put 300 thoughtful words in front of the teacher, every response different from every other!

Teachers now have a problem of abundance, not scarcity. They have to decide which responses to select, and why, in an environment of cognitive overload.

This is very hard work for teachers, especially novices, especially teachers who lack mathematical content knowledge, especially teachers who are hanging onto the school year by their fingernails.

We offer teachers lots of different support for discussions throughout our curriculum—both in print and digital activities—but our new discussion support for digital activities is first-of-its-kind and best-in-class.

Discussion Moments.

1. Student responses stream into the teacher’s dashboard.

2. A message appears: “Analyzing Student Responses.”

3. Shortly after, the message changes: “Open Discussion Moment.”

You click the message and see a classroom-ready discussion screen.

First, you see four student responses, each one authored by a student in the class, each one interesting on its own. This was not luck. Those responses were curated by a large language model at the direction of our curriculum experts. “Find three responses that capture the star in different ways,” our experts prompted the AI. “Responses that add anchors. That remove anchors. Find one response that might not capture the star.”

Next to those responses you see a question: “Which one is not like the others?” That question feels surprisingly well-matched for this math and for those student responses. This, also, isn’t an accident. Curriculum experts made that decision.

You click the right arrow and see a suggested narration for the Discussion Moment, narration which was authored, again, by our human authors for this particular problem, to help novices learn to facilitate productive discussions in math.

That’s a “Discussion Moment.” 👍

In the past, coaches, experts, and publishers have all asked teachers to . . .

• Select and sequence student responses.

• Construct a student-facing discussion resource.

• Lead the conversation.

Now we are asking teachers to . . .

• Lead the conversation.

In our experience, computers do quite well with the first two jobs while teachers obliterate computers at the work of leading a conversation, at connecting student ideas, at asking one kid what they think of another kid’s idea, at pulling ideas out of a kid who maybe doesn’t think they have ideas to offer. Discussion Moments delegate to humans and computers the best work for each of them.

Discussion Moments are different.

Lots of edtech companies are putting AI to work in lots of different ways. Discussion Moments are unique.

First, they’re designed to work through rather than around the teacher, during class rather than outside of class. They’re designed to support social interactions between students and teachers in the moment of instruction. This is the action.

Second, this is a classroom-ready resource. So many AI applications just output a ChatGPT-style resource. Lots of text. Several main bullets. Lots of sub-bullets. An emoji or two. And I am very sorry, but they are not useful in class. The teacher has to read all of that text, copy and paste and edit it, and then construct the student-facing resource all in the middle of class. That’s fantasyland, folks. At Amplify, we have, instead, created a one-click, classroom-ready resource.

Third, we’ve fortified these digital Discussion Moments with gallons of human expertise. Since December, I’ve worked with several of our curriculum experts—Casey Nelson, Brian Kam, and Tom Snarsky—and for every problem across several units of middle school math, they:

• Reviewed thousands of student responses to each problem.

• Identified thematic trends in the student responses.

• Decided whether or not those themes demand a discussion.

• Decided which of several discussion frames would be most appropriate, given those themes.

• Wrote an AI prompt specific to each problem to increase the odds that the large language model will curate useful student responses.

Most edtech companies would prefer to let AI lead this process from end to end, using the same prompt for every problem, even at the cost of the teacher and student experience. Meanwhile, we only ask AI to execute instructions and construct a resource. The nature of those instructions, the type of resource, and how it’s used—that is all determined by different humans and their expertise.

What do teachers and administrators think?

I ran a small-scale pilot of this feature last spring and kicked off a larger-scale pilot last week. A couple hundred teachers overall. I have never had an easier time recruiting teachers for a project than with this one. Every district math curriculum lead knows how challenging it is for teachers to lead discussions, and every one I asked was eager to support.

Two other examples of Discussion Moments.

Compare and Connect. We asked a large language model to locate responses that have one of a couple of important features but ideally not both. Then we constructed a Discussion Moment asking students to write a response that combines the best of both answers.

Critique, Correct, Clarify. Our curriculum authors noticed a frequent incorrect answer to a question. We told the LLM to watch out for it and frame it in a Discussion Moment where the class is asked to find value in the wrong answer before correcting it. Try to imagine what it does to a kid to hear their incorrect answer described as valuable.

Wait—don’t you hate AI?

I get why you might ask me that, but no. I think generative AI is perhaps the most overrated education technology of my lifetime; I don’t think the chatbot tutors or lessonslop generators are going to transform K–12 education. But I do think generative AI is neat. And look, I have tried to support discussion work with K–12 teachers for the last ten years in other ways, too. I have run in-person and remote PD. I have written math lessons and teacher supports for those lessons. I have sent nifty little customized email sequences tailored to teacher usage. None of those supports have been as promising as AI is here. None of them has moved the needle like Discussion Moments because none of them has been able to meet teachers in their moment of need, at the point of use.

That’s it. You can find Discussion Moments in Amplify Desmos Math next school year.

Next week, I’ll share more about the process of building Discussion Moments and how that process differs from the status quo in edtech.

Alex made this bet on LinkedIn:

I think by the end of [2025], it’s going to be hard to find an educator who doesn’t use Generative AI on a weekly basis.

I admired the specificity of that prediction, especially stacked against gauzy, non-falsifiable predictions like each of these, each one waving its hands and declaring 2026 “finally for real the year of the agent.”

I commented that I would take his bet and give him better than even odds. Not only wouldn’t we see anything close to 100% weekly usage, I said, but we wouldn’t see more than 80% AI usage of any type—casual, committed, or weekly.

Let’s check out both of those bets.

To Alex’s prediction of ~100% weekly usage, Gallup reported:

Only 32% of teachers report using AI at least weekly, while 28% use it infrequently and 40% still aren’t using it at all.”

Yikes! 32 is a lot less than 100, as far as numbers go!

Gallup also confirmed my prediction that fewer than 80% of educators would use AI in any capacity. You can triangulate Gallup with other sources too. Education Week has asked a national panel of teachers the same question every year for the last three years:

Which of the following best describes your current use of artificial intelligence-driven tools in your classroom?

The percentage of people who used AI at all in their 2025 survey (a little + some + a lot) was just 61%, almost identical to Gallup’s finding.

FWIW, Alex didn’t take me up on the bet I proposed. Nevertheless, I am bound by the Substacker’s oath to report to you that I would have soaked him.

What will happen in 2026?

AI use is certainly increasing among teachers, though that usage still seems confined to the dabblers in “a little” and “some.” The teachers using AI “a lot” are not exactly leaping off the axis. What kind of education technology is AI shaping into? High usage and high impact like slide software or an LMS? Low usage and low impact like the review games that teachers use every couple of weeks before a quiz?

I’m sticking with my predictions from early 2023 that AI will offer “quality-of-life improvements for teachers and students” but stop short of any “two sigma”-style changes to student learning or school staffing. Those predictions felt heterodox in 2023, but now I’d bet my house on any of them.

In 2026, you won’t see committed AI usage (“a lot”) rise above 10%. AI professional development will increase and you’ll see the dabblers swell in numbers—80% usage wouldn’t surprise me. But AI has not yet found its way to the core of teaching or learning. Chatbots and lessonslop still dominate the landscape. AI does lots of things faster and cheaper for teachers, but not enough things better. In 2026, you won’t find a room of educators where more than 10% of them would give up their slide software for MagicSchool. There are levels here.

What are the stakes of this bet?

More than I want to be right here (though I’ll take it) I want to be clear-eyed about the needs of teachers and learners, about the complex cognitive and social relationship they construct over the course of a school year, about the ways technology can and can’t support that relationship. The edtech industry is less sober than I have ever seen it, which turns your sobriety into a superpower.

My plan for edtech sobriety in 2025 involved substitute teaching monthly and helping a teacher out weekly. When 2025 rolled over to 2026, I immediately reset that goal. If you have perhaps misjudged the landscape of teaching and learning, if everything seems a bit blurry and you missed your predictions by 70%, I cannot recommend more highly that you shake your 2025 hangover by getting into a school. Try to teach something to someone who doesn’t really want to learn it, who doesn’t really think they can learn it. The experience will shape you and sharpen your senses and sober you up almost immediately.

Odds & Ends

¶ Eedi’s Chief Impact Officer Bibi Groot answered a question I raised in my review of their chatbot tutor study. Why did their chatbots help weeks later more than in the moment of tutoring? Groot speculates that the chatbot tutors drew more reasoning out of students than the human tutors, which supported a form of learning more amenable to transfer. I don’t know about that, but it sure is a thought! (I loved the graph above, indicating the human tutors eventually start cutting the chatbot tutors off, saying, “Hey—we’re done here.”)

¶ What innovation has re-energized this teacher?

“I feel like I’m coming home every day less exhausted, and I just have more energy to reinvest in the relationships with students,” Hesseltine said. “I’ve been teaching for over 20 years, so I really cannot stress this enough that this has made the biggest impact on education that I’ve ever seen.”

These are the kinds of results AI guys have been promising for three years now. So what is the teacher talking about and what stops us from getting it to more teachers?

¶ Just before the new year, Sal Khan wrote an op-ed in the New York Times considering employment in the age of AI. His solution to the problem of job loss is to pass a tin cup around to the companies profiting from AI and ask them to kindly donate 1% of their profits to programs dedicated to reskilling workers for AI fields. I can only imagine who Khan imagines would put those donations to work and lead the re-skilling effort.

The fund could be run by an independent nonprofit that would coordinate with corporations to ensure that the skills being developed are exactly what are needed. This is a big task, but it is doable; over the past 15 years, online learning platforms have shown that it can be done for academic learning, and many of the same principles apply for skill training.

In any case, the New York Times readers are absolutely scathing:

We have a mechanism for managing change equitably. It’s called the government. We don’t beg corporations for 1%. We tax them appropriately and use the tax funds to improve the lives of citizens and keep production and consumption within the limits that the biosphere has established.

¶ “Friction-maxxing” has entered the chat.

Friction-maxxing is not simply a matter of reducing your screen time, or whatever. It’s the process of building up tolerance for “inconvenience” (which is usually not inconvenience at all but just the vagaries of being a person living with other people in spaces that are impossible to completely control) — and then reaching even toward enjoyment. And then, it’s modeling this tolerance, followed by enjoyment and humor, for our kids.

¶ That article connected me to a recent post from Dylan Kane where he concluded:

That’s a tough reality of teaching: often the path of least resistance is the path of least learning.

1997

In 1997, education historian David Labaree observed that we love to fight about education and that our fights about pedagogy and policy really mask a fight about purpose. He identified several purposes for education, two of which have been locked in combat for the last several decades. From my perspective, we can confidently declare a winner and loser in 2025. Here’s Labaree:

From the democratic equality approach to schooling, one argues that a democratic society cannot persist unless it prepares all of its young with equal care to take on the full responsibilities of citizenship in a competent manner.

The social mobility approach to schooling argues that education is a commodity, the only purpose of which is to provide individual students with a competitive advantage in the struggle for desirable social position.

2025

The dominant education trends of 2025—including AI, personalized learning, increasing student absenteeism, education savings accounts, school choice, microschools, the kidsmaxxing of Alpha School and its partners on X—all indicate an American public that is giving up on democratic equality and is trying, instead, to maximize the social mobility of their children however they can.

Traditional public schools are one of the few remaining places where we gather very different people together in the same room to learn and be nice to one another. If you’re interested in fostering democratic equality, that work is extremely important and it isn’t cheap. Teaching students to learn math and learn to be nice to one another across their widening social, academic, and financial differences represents a huge challenge for teachers. Additionally, learning math and learning to be nice often takes more time than learning math alone, because learning more stuff generally takes more time than learning less stuff.

In 2025, we watched more and more parents say, “We really care a lot about the learning math part, but the costs of fostering democratic equality seem quite high.” Consequently, we watched more and more parents opt out of the relative heterogeneity of public schools and enroll their students into more homogeneous learning environments.

Not every parent group, though. Abigail Francis and Joshua Goodman studied enrollment rates in Massachusetts last year and found that enrollment declines in public schools were largely driven by wealthy parents. “The highest income 20% of districts have lost more public school students than the other 80% combined,” they found, “with these lower income districts having largely recovered.”

I think this is a dismal turn for our democracy, one that has been decades in the making, one that seems likely to continue in 2026. As the project of democratic equality provides fewer and fewer material benefits to people outside of schools—e.g. basic goods are very expensive; wealth inequality has only grown over the last 30 years; jobs are hard to find even for people who did all the things grown-ups told them to do like go to college and learn to code—people have, I think quite expectedly, prioritized that project less and less inside of schools.

Democratic equality has not lifted all the boats as promised. The ship of state is taking on water. People with the means to secure a lifeboat for their kids are doing exactly that.

2026

For 2026, as a citizen, I am most interested in projects that are committed to social democracy, committed to the truth that we have enough resources for everyone to have a lifeboat and more besides to fix the damage so none of them are necessary.

For 2026, as an educator and technologist, I am most interested in supporting democracy in schools, in working with teachers towards classrooms where every student feels valued, where every student feels like their ideas are needed, where every student feels like this class and its learning would have been less had they been absent today. I have several projects planned for 2026 and each of them will try to make it much easier than it is now for teachers and students to learn together.

The Headline

The Reality

The Study

AI tutoring can safely and effectively support students: An exploratory RCT in UK classrooms

What They Did

The researchers embedded three kinds of support inside Eedi, a math software platform1.

• Support #1: Static pre-written hints.

• Support #2: A chat interface for accessing support from a human.

• Support #3: A chat interface for accessing support from an LLM, with a human approving the messages.

The study was nicely controlled. Every condition got tested against the other. Support #3 is the focus of the study. “In sessions with LearnLM, a supervising tutor reviewed each message that LearnLM drafted. They could either edit the message, completely re-write it, or approve it without any changes.”

Their outcome measures were whether students corrected a mistake, resolved a misconception, or transferred their knowledge. (See p. 9 for more precise descriptions of those outcome measures.)

Can I See a Picture?

What Happened?

Both of the dynamic supports clobbered the pre-written hints, as you’d probably expect.

There wasn’t much difference between “human alone” and “human + AI” on mistake remediation or misconception resolution. The big headline: the human + AI saw a 5.5% increase over a human alone on knowledge transfer. The study reports the odds that this is a legit finding at 1.3.

Finally, we directly compare the two tutoring conditions. We estimate that receiving support from LearnLM improved a student’s odds of success by a factor of 1.3 [0.9, 1.7] relative to human tutors, corresponding to an ATE of +5.5% [–1.4%, +12.4%]. Based on this posterior distribution, we find a strong probability (93.6%) that LearnLM elicited greater knowledge transfer than human tutors alone.

If, like me, you aren’t particularly familiar with the Bayesian methods here (the odds ratio) and if you are likewise concerned that the confidence interval contains 0 and the odds intervals contains 1, you can read this response from one of the study contributors, who seems to say, “Yeah, the odds aren’t 100% but they’re a lot better than chance.”

The study stored the original AI responses and the edits from human tutors (a fun source of data) which led to the finding that 76% of the LLM responses were accepted in full by the human tutor.

LearnLM proved to be a trustworthy source of pedagogical instruction, with the supervising tutors approving over 76% of its messages without changes or with only minimal edits (changing one or two characters; e.g., deleting an emoji).

Reasons to Be Happy

• It’s a responsible study. Humans supervising the LLM. Lower risk than just setting the LLMs onto the kids directly.

• “No differences” is still good. I think we can chalk something of a win for AI even when there isn’t a significant difference between the human tutor and human + LLM. The LLM operates under different constraints than a human. It can work faster and support more students simultaneously. The mere fact that there wasn’t a degradation in student learning is reason enough to be happy.

Reasons to Be Real

• It’s a small practical effect in a short intervention. The total time here was seven hours—one hour per week over seven weeks—for an increase of 5.5% on some proprietary measures. I would put this study in the basket of “point solutions that don’t congeal in any obvious way into a larger plan for school improvement.” Point solutions aren’t bad but we need to be real about the scope of the intervention here.

• This study is trying to do the wrong thing the right way. If the best tutoring interventions have kids a) working with trusted community members (parents, paraprofessionals, etc) b) in person for c) multiple times per week, d) doing work that is aligned to their curriculum, this intervention has none of those hallmarks. In general, getting tutored by a human through a text messaging interface shaves off a huge amount of the value of getting tutored by a human while also playing to a text chatbot’s strengths. It’s as if we asked Michael Bublé to play the kazoo and then supervise a robot purpose-built to play a kazoo. I’m not sure we’d be surprised by the results. I’m also not sure we’d feel good about saying “AI+Michael Bublé Match Quality Of Michael Bublé Alone.”

Reasons for Surprise

• It’s honestly quite strange to me that the Human + LLM showed its strengths with “Knowledge Transfer.” I would have expected to see the greatest difference on follow-up tasks that were closer to the tutoring moment. Wouldn’t you? Kids didn’t fix their answer on the next try (“mistake remediation”) or the try after that (“misconception resolution,” I guess) at different rates between the human and human+AI conditions. But apparently if you look beyond those moments, to a question in a different unit the next week, the effect of the AI really kicks in?

• This isn’t what’s commonly called “Knowledge Transfer.” Doing the next topic is doing the next topic. Knowledge transfer is typically understood as taking your learning on a topic and transferring it to a different context. i.e. “Okay you learned about slope in the context of an airplane landing. Can you calculate it in the context of the unit price at a grocery store?”

• I’m still surprised by this. If I were cynical, I’d wonder if the researchers found no significant differences on those first two much more intuitive measures and then went looking around at other measures, a frowned-upon process known as p-hacking. But thankfully I am not a cynic—a sweetie, in point of fact—and I will just process my cognitive disequilibrium on my own.

Implications

In these kinds of text-based tutoring environments, it seems possible AI could do some good or at least not do harm. Even if the LLM did fine in 76% of exchanges, though, I suspect there is some real frustration in the long tail of the distribution. For example, one human tutor said they had to overrule the LLM’s tendency to question students to death:

Tutors often found it necessary to step in when LearnLM’s Socratic questions, while pedagogically sound, persisted longer than a student’s patience. One tutor described a common scenario where “[LearnLM] will go, ‘Okay, you’ve got the answer. Let’s dig a little deeper about why you’ve got that answer.’ And the child is just like, ‘No, I’ve got it. I know what I’m doing. Can I go now?”’ (T1).

If the questions persist past their usefulness, I don’t think we can call them “pedagogically sound.”

Odds & Ends

¶ Over on LinkedIn, Shelbi Cole posted the image above and wrote:

The only reliable way to create strategic groups for Tiers 2 & 3 math instruction is to understand what children are doing.

The kid gets one out of six problems correct for a score of 17%. But “one out of six problems correct” is a pretty limited understanding of the child’s understanding and actively unhelpful for prescribing an intervention for the child. That’s a challenge for digital systems that reduce student thinking down to a binary correct / incorrect mark and that don’t make room for student handwriting.

¶ I was a guest on the last EdTechnical podcast of 2025. Libby Hills and Owen Henkel have a great vibe, know the space backward and forward, and keep a brisk conversation. We chatted about discovery learning v. explicit instruction and about AI Guys, how they’re made and how they limit their own effectiveness in edtech.

¶ Dylan Kane describes High Accountability Teaching in a post I liked quite a lot. Here’s the tease:

The core of high accountability teaching is to get as many of the swingers on board as possible.

Lately, I think a lot about “mattering” in classrooms. Do kids feel like they matter in their class? Do they feel like their ideas matter? Do they feel like the work they are doing matters? When kids complete work in class that a) isn’t seen by other kids, b) isn’t seen by grownups, c) isn’t commented on, d) doesn’t count for a grade, when they start to wonder, “Does this work actually matter to anyone?” that spins up a pretty negative cycle for learning. Dylan’s post added a lot to my thinking here.

¶ University of Washington freshman Luran Yang wrote a beautiful op-ed in the Seattle Times responding to the near-constant use of AI on her campus. “Not all hope is lost, however — I’ve been in one class, and only one class, with a near-zero rate of AI usage.”

¶ Education Week reports on Austin ISD’s use of artificial intelligence to support the annual creation of their master schedule, a task that historically requires lots of manual shuffling of teachers and classrooms and students.

… it saved [principals and their scheduling team] at least 50 hours. And the feedback that we got from those schedulers was that it increased their job satisfaction. They were so excited they got their summer back. They felt like they could actually manage master scheduling, and the master scheduling wasn’t managing them.

I don’t hold up “time savings” as a first-order goal for teachers, administrators, or schools. There are plenty of ways to save time that result in a worse education for kids. But I am much more optimistic that AI can support back-office operational tasks like this rather than the in-classroom relational work of teaching.

• Letting you know how the AI + education debate in Washington, D.C. went on Monday a/k/a 🤓 nerd fight night 🥊.

• Sharing two of the fantastic ideas for teaching slope you all shared in response to last week’s newsletter.

On Monday, four of us—

1. Shanika Hope, Google

2. Alex Kotran, AI Education Project

3. Jake Tawney, Great Hearts Academies

4. Me

—argued about the motion, “Maximizing School Improvement by 2035 Means Integrating AI into Classrooms Today,” in Washington, DC, at an event hosted by the American Enterprise Institute. We all had three minutes for our opening remarks. Here is what I prepared.

Look—I’m not made of stone. I love the motion and wish it were true, this idea that we made a massive technological breakthrough three years ago and now we can improve schools like never before. I wish it were true. There are a few reasons why I don’t think it is.

The first is the work we have asked schools to do in 2025. We have asked schools to help kids read, write, do math, and be nice to each other all at a time when hormones are flying around the room like pinballs, in a country that has some of the highest child poverty rates of the developed world, at a time when kids are walking to school scared that they or their friends or their loved ones are going to get jumped out on by some dude in a mask dragging them to God knows where. It’s a hard job at a hard time. I tried it last week. I tried to teach slope of a line to eighth graders in East Oakland, wonderful kids in the Fruitvale neighborhood. Ask me how well that went. Kids have a lot more on their mind than math right now.

The second is to understand the tight resources we have given schools to do that job. Financial resources are tight, obviously, but I’m talking also about time. We give schools 180 days to turn 3rd graders into 4th graders. We give schools four days for teachers to get out of class, get together, and learn something new. And perhaps most importantly, administrators have limited reserves of social capital. You can’t go to your school board, your parent community, ten times a year and say, “Okay, we’re going to try something new, something big, everybody get on board.” You can do that once, maybe. Opportunity costs are high. Saying yes to one thing means saying no to ten other things by definition. So that one thing needs to have absolutely bulletproof evidence that it’ll help your school help kids learn to read, write, do math, and be nice.

So my third reason is that in 2025, after three years of trying, the evidence that AI can help schools do that work is extremely weak. I know you have seen the same happy headlines I have or the profiles of schools doing something novel with AI. But scratch even a millimeter beneath the surface of those studies and stories and you’ll find some very weak evidence. You’ll find weak controls, where the control group gets nothing, as it should, and the experimental group gets AI, as it should. But they also get something extra—extra time, extra tutoring, extra rich parents. Elsewhere, you’ll find positive results for AI but they’re in a basket full of point solutions that don’t congeal in any obvious way into a school improvement plan. Still other studies run the other direction, like the MIT study this summer, where kids who used AI to support their essay writing were effectively de-skilled, experiencing diminished cognitive activity and unable to recall the arguments they’d made, compared to students who didn’t use AI.

For those reasons—hard job, limited resources, high opportunity costs, weak evidence in favor of AI, my recommendation is we wait. We take a beat. We take the next bus. We wait for stronger evidence and then move. For now, if you want to maximize school improvement, it’s an easy call to wait rather than integrate AI.

My comrade Jake Tawney argued that these tools would inhibit the formation of novice learners and novice teachers alike. In their arguments for the motion, Alex Kotran and Shanika Hope argued that:

• Kids will need AI skills for full US economic participation

• AI is being used outside of school, often poorly, and schools need to help kids learn to use it well

• AI will enable teachers to do better work than ever before, planning, differentiating, personalizing, preventing burnout, etc.

The debate was decided by a vote from the audience before and after the debate. The winner was whichever side most increased their share of the vote.

We won, which was fun of course, and the conversation kept everyone dancing, trying to stay on message, trying to answer the question we wished we’d been asked without seeming too obvious about it.

Jake and I counter-argued that many of the claims about the value AI offers teachers or students were dreams about the future rather than descriptions of the present. It is also a particularly unfortunate time to suggest that “we know what kids will need for economic participation in ten years,” given we are in the weakest labor market that software development has ever seen, ten years after telling every 12 year-old they needed to learn to code. That effort has gifted the private sector an oversupply of software developers who now have double the unemployment rate of art history majors.

If I had to guess what persuaded the audience to join us against the motion, it was our focus on “maximizing school improvement today.” In my closing argument, I said there were a bunch of motions I thought we’d all agree on, including the need for responsible experimentation, the need for dreaming expansively about what technology might offer kids. But maximizing school improvement today means taking seriously what a school most wants to improve, which is a kid’s ability to read, write, do math, and be nice, and applying the research we know works today. I suspect that approach gave some audience members permission to think AI is quite neat but also that 2025 is not the year that AI comes off the bench to maximize school improvement.

Mathematical Mailbag

I expressed difficulty last week keeping students thinking on a conceptual level about slope when the operations (“subtract, subtract, divide”) are so close at hand. Here are a couple of your very interesting responses.

Linda, offering an anchor activity that students can refer back to time and again as they try to make sense of new, related ideas:

1. I asked [students] if they realized that if you build stairs in your house that are too steep, it would be “illegal” because it could be dangerous. And that less steep stairs can be inefficient because they might take up too much space in a house.

2. I then gave each pair of students 12 identical Lego bricks.

3. I asked them to make 3 stair cases, using 4 bricks each, of different steepness.

4. After checking their stairs, I had them lay the 3 cases sideways on paper and had them trace it.

5. I then had them determine and record the rise and run of each of the stair cases. I checked those too.

6. Finally, they were to connect the top to the bottom and draw that line with a ruler. At that point, I asked them to record, “What do you notice about the lines you just drew and the slopes of the stairs?” We had conversations including the fact that some people’s stairs sloped downward to the right and others up. So, we also made that connection to positive and negative slope. Some noticed that larger slopes signified steeper stairs (and lines). In following days, when students were working on slope we would refer back to our “Lego stairs” to conceptualize what was going on.

Michael Hayashida, via email, illustrating how to help students understand the need for the measure known as slope:

Here’s something that seems to work in my freshman (somewhat remedial) math classes. All the kids have portable whiteboards.

“Draw me a set of axes and then draw a line that’s really steep.”

“Hold up the whiteboards - anyone have a line with the same steepness as someone else?”

“Ok, draw me a line that’s only medium steep.” They compare again.

“Ok, for mathematicians, this is a problem. They want precision - so that everyone draws a line with the same steepness when I give the instruction. To do that, you need to be able to put a number on steepness. Like... if I call for a line with a steepness of 5, then everyone draws a line with the same steepness. For the next minute, talk to your partner and see if you can come up with a way to measure steepness with a number.”

We share out. Sometimes kids will come up with an angle measurement method - kinda cool.

“Ok, this is actually a solved problem - mathematicians have already agreed on a way to measure steepness. Check this out - everyone put your marker on a lattice point - doesn’t matter which one. Now go over 1 and up 2 and put a point there. Do it again. Do it again. Connect the points. Hold them up - does everyone have lines with the same steepness? Ok, mathematicians call this a steepness of 2.

<Do the same thing, but with a slope of 0.5. Then 2.5.>

“This is how mathematicians decided to measure steepness - it’s what the y is changing by as x goes up by 1. They use a different word than steepness though - they call it slope”.

In the past, I emailed you every Wednesday about whatever interested me most that week. My new plan is to email you only on Wednesday, and only when something has interested me about both a) math education, and b) edtech. You will never ever receive this newsletter on any day other than Wednesday and those Wednesdays might be months apart. That’s the Dan Meyer promise.

First, some plugs.

Here are a couple of time-sensitive announcements.

A debate about AI in education in Washington, D.C. Next Monday, December 8 at 6:30PM Eastern, I will be debating the motion “Maximizing School Improvement by 2035 Means Integrating AI into Classrooms Today.” I will be arguing against the motion (surprised??) with Jake Tawney, a leader in a classical school network. Jake and I will debate Alex Kotran, founder of the AI in Education Project, and Shanika Hope, who works on education initiatives at Google. The American Enterprise Institute is hosting the debate and I hope all of us will walk away a little smarter. You can attend in person or watch online.

A prediction contest about AI in education. The EdTechnical duo of Libby Hill and Owen Henkel are inviting you to put down some concrete predictions about AI and its future impact on education. There are five tracks with a top prize of $2,500 in each track. I am happily serving as a judge in the “Teaching Profession” track. We’re inviting predictions for the question: “By the end of 2028, what percentage of non-interpersonal teacher activities (lesson planning, grading, and parent communication) will teachers routinely delegate to AI systems?”

Yesterday’s humiliation.

Richard Rohr prays for one good humiliation daily to help him see the difference between how he thinks things are and how things really are. I beg people to let me take over their classes for roughly the same reason.

It’s dishonest work—building tools for teachers. Figma designers do their design work in Figma. Cursor devs develop software with Cursor. Google employees use the Google Suite. They all use their own tools regularly. They all regularly feel the same pain as their customers. For a few good reasons and a lot of bad ones, it’s hard to find edtech developers “dogfooding” their own tools1. This is why our team here at Amplify ritualizes classroom visits. This is why I beg.

I was teaching kids to calculate slope from two coordinates yesterday here in Oakland, CA. Operationally, this is very easy. Subtract, subtract, and divide. Conceptually, this is a topic that’ll bloom for at least the next five years. What does the slope mean? Why do we divide the vertical change by the horizontal? Where do the graph and the algebra connect? What happens when the horizontal change becomes very, very small? What is the slope of the slope?

The core challenge.

How do you “press” 30 kids to develop important conceptual knowledge when the easy operational knowledge is right there? I found myself doing a lot of desk work, trying to figure out what ideas students were bringing to the table, offering them questions to help them develop their intuition about slope.

• Okay which quadrant is this point in? Is this other point to the left or right? How do you know?

• Here are some points. Pick a couple of them you think would make a downward sloping line.

• How high did I have to climb from this point to this other one?

• Let me offer a wrong answer. How can we know it’s for sure wrong?

• Etc.

It would have hastened the whole interaction to direct them towards the easy operational knowledge. Subtract, subtract, and divide. I don’t know how to shortcut the conceptual work here. It’s possible to do some of it from the front of the class, but most of it requires serious, individual engagement with kids and their ideas. I don’t know how to do that across 30 kids. And when you turn that work over to a computer, it will happily encourage the kids to subtract, subtract, and divide.

I need new ideas for doing conceptual work at classroom scale. Let me know what you have on hand.

Odds & Ends

I have eight months of great links to catch you up on. Let’s start with five.

¶ I appreciate the skeptical posture Michael Pershan adopts towards just about everything. Recently, he directed it towards the overly tidy explanations our pundit class has offered for America’s declining test scores.

Here’s the situation: Americans are getting dumber...well, mostly not. But our lowest performing students seem to be losing ground. And, simultaneously, our adults. Some international tests show a similar decline happening in other countries. On other exams, America is on its own. What gives?

¶ Similar from Freddie deBoer: Was the United States Once a Global Leader in Educational Metrics? Have We Fallen From Those Lofty Heights? No and No

So here’s the nut of the whole thing: commentators often point to the recent decline in test scores (post-2020) as proof of a unique American failure. This ignores the global context: almost all developed nations have seen declining scores, especially in the context of the Covid-19 pandemic, and in fact in many countries these declines have been more pronounced than in the United States.

¶ Justin Reich’s podcast The Homework Machine concluded its run during my hiatus. It was extremely well-reported. Great interviews. Great perspectives on teaching, learning, and technology. In its final episode, a San Francisco English teacher named Sara Falls is asked why AI isn’t really helping her and she proceeds to describe the core challenge of teaching in a way that still haunts me months later. You should listen to her tone in that moment, but if you must, you can read what she says:

… this kid who just brought me this candy cane … this sweet kid … I’ve struggled … the fact that she brought me a candy cane is a big deal because at the beginning of the year she hated me. You know … and I’ve stuck with it … I’ve stuck with her and she just brought me a candy cane. This is the work.

¶ While I was out, OpenAI, Anthropic, and Google, all released “study mode” versions of their chatbots. These new versions won’t just throw you the answers when you ask. They’ll make you work for those answers. (The marshmallow test goes to college!) I enjoyed reading Mashable’s Chase DiBenedetto’s reports from test-driving each one, especially how the directive to not give the answer ever can drive a student to madness:

But here’s the unsettling part of using Claude: As conversations continued, it increasingly felt like I was trapped, in the sense that lessons seemed to go on forever. Because Claude wants to really make sure it knows where your head is at, the bot doesn’t provide you a clear stopping point for the lesson, sometimes making conversations hard to follow, and the learning experience frustrating. During our math session, Claude fervently lauded me when I got the answer right, but sent me approximately 100 follow-up questions. Even when I asked if we were done with the problem — AKA can I go home now? — it gave me more work to do.

¶ Related: I loved reading RAND’s survey of students and their use of AI and I don’t think I saw anyone report on it. One interesting question asked students “so what tools are your teachers recommending?” and then “what are you actually using?” I went ahead and plotted both sets of data simultaneously and you can get a preview of the popularity of these study modes. The max teacher recommendation was for Khanmigo—a Socratic tool. Kids went ahead and used vanilla ChatGPT anyway.

• Keeping me focused on what’s going on in the world of education and technology.

• Sharpening my own ideas and fluency in talking about education and technology

• Developing a community (that’s you!) that makes me smarter through your comments and feedback.

But writing weekly consumes a non-trivial amount of my time and I have a new opportunity at Amplify to build something I have wanted to see in the world of edtech since pretty close to November 30, 2022. So I’m going to reallocate my time from here to there and I plan to reallocate time back whenever that’s possible.

The AI Rapture Ain’t Nigh: What To Do When You Stop Waiting

I’m going to leave you with a video of the 15-minute talk I gave three days ago at ASU+GSV, the largest gathering of edtech investment and AI hype anywhere in the world. Whether the talk is any good is for other people to decide but I feel convinced that no one else could have given it. It combined:

1. A perspective towards kids that is grounded in their assets rather than their deficits.

2. An understanding that the work of teaching is multi-faceted—cognitive and social simultaneously.

3. A vision for mathematics that is creative and connected.

4. A blueprint for technologies that can increase the odds of kids experiencing #1-3 above.

Plus:

• A rebuke of all the people in edtech whose work diminishes kids, teaching, or math. Even if I am not writing this newsletter weekly, please know I am still wishing you the worst of luck.

The talk is distinctly mine and I’m happy to leave it as my last post for I don’t know how long.

Until I’m Back

• Feel free to get in touch by hitting reply to this email or going direct to dmeyer@amplify.com.

• For AI & edtech, please read:

  • Audrey Watters

  • Benjamin Riley

  • John Warner

• For math ed, please read:

  • Dylan Kane

  • Michael Pershan

  • Robin Kubasiak

“IDK” I told Khanmigo in response to its every gentle, Socratic nudge, and eventually Khanmigo said, effectively, “Okay here is the answer. Does it make sense?”

Human tutors aren’t so easily manipulated and if you believe AI chatbot tutors should someday approximate human tutors, these are the problems you have signed up to solve.

As is frequently the case, some Khanmigo product changes followed my post critiquing Khanmigo. Now, when I spam Khanmigo with IDKs, Khanmigo effectively stonewalls, saying, “Okay look, here are some resources to watch or read. Quit wasting my GPT tokens with your nonsense.”

This is an interesting development! The website JoySchooler is another platform that refuses my manipulation, saying, “Let’s take a break,” and then disabling the chat box.

My opinion is that these are both interesting developments and worth sharing with you, but they both still look a lot like hell. It still seems like deeply alienating work, trying to build a skilled human teacher from a sack of LLM prompts, one with low odds of success. With every new step towards your destination, you realize it’s two steps farther than you thought. Every new line of your prompt risks contradicting a previous one. You write a new line to meet a new student need and accidentally unmeet an older one.

And let’s say you manage to create a chatbot that says exactly what a skilled human tutor would say at every turn in a tutoring interaction. Will we then see skilled human tutor results from the chatbot or will we find that the real liability here wasn’t that the chatbots were saying the wrong things but that kids don’t care what chatbots have to say? These are the questions you have signed up to answer.

Featured Comment

Too many to share from last week’s post on teaching the long cut, but I liked Kristi Peterson’s description of “skipping the math” a lot:

I love using the "long cut" for graphing linear equations. When working with the Slope-intercept Form, most resources I've used go straight for the short cut of plotting the y-intercept and using the slope to plot the points of the line, but this is usually meaningless to the students. Instead I require that they use a table, then identify the slope and intercepts. Usually after about 20 linear equations, they see it for themselves. Then they think they can trick me by "skipping the math" and just plotting one point and using the slope -- jokes on them :)

Also Bryan Kerr on why so many software developers find ideas like a “Personalized Netflix for Education” so seductive:

I think Netflix-i-fying content may work well for those with high intrinsic motivation. I'd love it if I could get customized videos to help me vibe code something that helps me solve an important problem. But edtech companies may be putting the cart before the horse as their designs seem to take for granted that students already see the value in what the software delivers to their screens. Maybe the developers just assume kids are as highly motivated as they were back when they devoured YouTube videos learning to code.

Odds & Ends

¶ David Wiley is a smart guy but he’s trying to do the wrong thing the right way by teaching an LLM about pedagogical content knowledge. How … can I put this more simply? Kids do not seem like they want to talk to LLM chatbots about academic content, no matter how well they are trained. The oracle might contain all knowledge and know the answer to every question, but none of that matters if she is surrounded by dragons and an acid moat. The medium inhibits the message.

¶ More from Khan Academy. First, Sal Khan has an interview with CNBC where he says:

The tools are exciting and we’re investing a lot in this front but in most cases it’s about how you integrate it with the classroom, how you help the teacher hold the students accountable, engage them, and then the students will be off to the races.

I think that’s exactly right and I am very curious how Khan Academy intends to “help the teacher.”

¶ Recent predictions from some edtech luminaries. Bill Gates was on Jimmy Fallon’s show and said:

The era that we’re just starting is that intelligence is rare, you know, a great doctor, a great teacher, and with AI over the next decade, that will become free, commonplace. Great medical advice, great tutoring.

I just want to note the sleight-of-hand that turns a “great doctor” into “great medical advice” and a “great teacher” into “great tutoring.” Which people get which of those resources is going to be a matter of intense debate over the next decade and we must not pretend they are the same thing.

Another interesting prediction. Sal Khan recently told the National Governors Association:

I think the best scenario is that in 2034 if you are a teacher, you can spend 90% of your time on student-facing tasks.

Wait—that’s what he said about 2024!

Looking ahead to 2024, I see generative AI tools cutting 90% of teachers’ admin tasks, creating more time for student interaction.

AI was supposed to give teachers 90% of their time back at the end of 2024, Sal! Now they have to wait ten more years? What is going on?

There are lots of contexts where missing your forecasts by 10x would immediately discredit a person, but edtech is apparently not one of them. In edtech, you will never get asked, “So what happened there? What’d you miss? Did you misunderstand the ed or the tech or both?” In edtech, it seems we’re all just having a bit of fun.

¶ Our edtech luminaries are softening their predictions for generative AI. The biggest generative AI IPO underperformed expectations. Microsoft is canceling data center leases. Salesforce’s big AI play isn’t delivering on revenue expectations. There is increasing evidence inside and outside of education that generative AI is not anywhere close to delivering on its promises. I still think this technology is neat but the sooner we grip the possibility that it isn’t much more than that, the sooner we can get back to work.

¶ Tressie McMillan Cottom names generative AI a “mid technology” in the New York Times:

That tech fantasy is running on fumes. We all know it’s not going to work. But the fantasy compels risk-averse universities and excites financial speculators because it promises the power to control what learning does without paying the cost for how real learning happens.

¶ I’m a huge fan of the CourseKata team and their joyful approach to data science and statistics. They’re offering a fellowship for high school teachers that includes free PD and curriculum. Priority is given to early applicants, so I encourage you to check out the opportunity.