What Happened When My Son Spent a Month at an AI School
The first thing I noticed was how quiet it was. Not library quiet, but the kind of quiet that made me check whether my watch was still working. My son Leo, nine years old and generally incapable of sitting still for more than four minutes, was staring at a screen with a set of headphones on, occasionally scribbling on a whiteboard. Next to him, another kid was working on a different subject entirely. Across the room, a grown woman—called a “guide,” not a teacher—sat with a small group, talking in low voices about a book they’d all read.
I’d signed Leo up for a two-week summer program at a school that used AI for core academics. Alpha School, and others like it, have been popping up in cities, promising that children can learn math, reading, and science at their own pace, guided by software that adapts to every mistake and every correct answer. No more whole-class lessons where half the kids are bored and the other half are lost. No more homework packets. Just a computer that knows exactly what your child doesn’t know, and won’t move on until they get it.
That was the pitch. I was skeptical, mostly because I’d spent eight years watching teachers try to do exactly that with nothing but patience and a stack of differentiated worksheets. Could an algorithm really replace that? The answer, I learned, is more complicated than yes or no.
On the first morning, Leo came home and said, “School was easy.” My husband and I looked at each other. Leo has always been a decent student but someone who rushed through assignments to get to recess. At his regular school, his math homework was a battle most nights. But this AI program kept giving him problems until he’d mastered the skill, supposedly. So why was it easy? Because he was starting from a lower grade level. The software had tested him and found gaps I didn’t even know existed.
That stung a little. I thought he was doing fine. His report cards said he was “meeting expectations.” But the AI didn’t care about report cards. It asked him to solve a fraction problem and then watched him flounder. So it backed up two steps, gave him a refresher on equivalent fractions, and watched again. It didn’t sigh. It didn’t say “we covered this last year.” It just waited until Leo actually got it, then moved forward.
By the end of the first week, Leo had completed about three weeks’ worth of math lessons. Not because he was suddenly a genius, but because the software didn’t let him skip ahead or fake understanding. He told me one evening, “The computer doesn’t think I’m good at math. It thinks I’m okay at fractions but bad at measurement.” Being called out by a machine, in a weird way, made him more honest about his own skills. He knew it wasn’t judging him, just tracking him.
The most interesting part wasn’t the screen, though. It was the guide. At Alpha School, there are no classroom teachers in the traditional sense. Instead, there are adults trained to ask questions, manage conflicts, and support projects. When Leo got frustrated staring at a problem, the guide didn’t help him solve it. She said things like, “Which step feels stuck?” or “Talk me through what you tried.” She wasn’t delivering content. She was coaching him through the process of learning.
That flipped my understanding of what a teacher is supposed to do. For years, I’d thought of teaching as explaining things clearly. But in this model, the AI handled the explaining. The guide handled the human stuff: noticing when a kid was about to cry, mediating an argument over who got to use the good markers, and crafting enrichment projects that no algorithm could design. One afternoon, Leo and a group of kids were building a model of a city out of cardboard, trying to figure out how to make a water system flow downhill. That was the guide’s job—to set up a problem messy enough that the kids had to work together, argue, and try again.
I found myself wondering whether my son was actually learning anything, or just getting better at using an app. So I asked him for a test. Give me a percentage. He said, “Fine, 17% of 200 is 34 because point 17 times 200 is 34, but also because 10% is 20, 5% is 10, and 1% is 2, so 10+5+2=17% and 20+10+4=34.” Then he went back to playing with cardboard.
Okay. He learned something. But here’s what concerned me: the social aspect. His regular school has twenty-four kids in a class, and Mr. Davis has to herd them through reading groups and science fairs. At the AI school, there were only fifteen kids in the whole room, and they weren’t all doing the same thing at the same time. Some were on math. Some were reading. A few were working on a history project. There’s a real benefit to that flexibility, but I also worry about kids who need more structure. Leo thrives with a little chaos. A shy kid might not get the same push to speak up.
I asked one of the guides about this. She told me that they deliberately schedule collaborative projects every morning, before the AI work, so kids get time to argue, negotiate, and figure out how to share space. She said it’s non-negotiable. “The software handles the facts. We handle the feelings.” That felt a little trite, but watching Leo negotiate with another boy over how many windows the cardboard library should have, I realized there was something to it. The kids weren’t passively absorbing lectures. They were actively making decisions, and sometimes the AI wasn’t involved at all.
The other thing I noticed was the parent-facing dashboard. Every day, I received a list of skills Leo had worked on, how much time he spent on each, and his accuracy rate. I could see that he struggled with division on Tuesday and nailed it on Wednesday. I could also see that he spent nineteen minutes staring at a video about the solar system, which I later learned was because he’d paused it to draw the planets on his whiteboard. That level of detail was both fascinating and slightly unsettling. I felt like I was micromanaging his education from my kitchen.
I started checking the dashboard more than I checked his backpack. It became a little obsessive. His real school sends home a weekly email with a general update and a few photos. This was different. I knew exactly what he didn’t know, and I had to resist the urge to drill him at dinner. I had to remind myself that reclaiming fraction knowledge was the point, not making him feel like I was spying.
After four weeks, Leo went back to his regular school. He had to readjust to sitting in a desk, waiting for other kids to finish, and raising his hand. The first week back was hard. He kept asking why the teacher was “talking so much” and when he could work on his own. But within a month, he settled in. His math teacher sent a note saying his number sense had improved, and he seemed more confident when volunteering. I don’t know if that’s because of the AI school or because he’d just matured over the summer. Probably a bit of both.
The experience made me think about what we actually want schools to do. We want kids to learn skills, obviously. But we also want them to learn how to sit with frustration, how to ask for help, how to work with people they don’t like. Those things aren’t easily automated. AI can track a math problem, but it can’t see when a child is trying not to cry. It can’t notice that a shy kid hasn’t spoken in two hours. That’s why, in the better versions of these schools, the human adults aren’t teaching content. They’re teaching attention.
I still don’t know if an AI-heavy school is the right fit for Leo long-term. He’s a social kid, and I worry about too much screen time. But I also watched him light up when he solved a hard problem that the computer had specifically calibrated for him. That feeling of “I actually get this” is hard to manufacture in a classroom of thirty. Maybe the future isn’t either/or. Maybe it’s a blend. Leo learned more math than I expected, and he also learned how to build a cardboard city with a complete stranger. I’m not ready to pull him out and enroll him fully. But I am ready to stop assuming that “school” has to look like it did in my childhood.
He still asks me sometimes why we don’t just let the computer teach him everything at home. I don’t have a great answer yet. Yesterday, he was trying to explain a new video game to me, and I got lost about thirty seconds in. He looked at me with the patience of someone who has watched a dozen videos on how to simplify fractions. “Don’t worry, Mom,” he said. “I’ll adapt the lesson to your level.” I’m still thinking about that.
Please indicate: Thinking In Educating » What Happened When My Son Spent a Month at an AI School