We Tried an AI School Like Alpha School: What Parents Should Know
The first morning we walked into the learning space, I wasn’t sure what to look for. The room was bright, with glass walls and clusters of wooden tables, but there was no chalkboard at the front because there wasn’t really a front. Kids sat in small groups, headphones on, working through screens. One boy was stretched out on the floor with a laptop propped against a beanbag. A woman in her twenties—called a “guide,” not a teacher—knelt next to my son, who had his hand frozen over the keyboard. She didn’t point at the screen. She just whispered something to him and tapped his shoulder twice. He typed an answer, and the screen changed color, and he moved on to the next problem.
We had signed our son up for an AI-based micro-school, the kind often compared to Alpha School. If you haven’t seen the model: kids spend roughly two hours a day on AI-driven academic software for math and reading, working at their own pace through a sequence of mastery-based goals. Then the rest of the day is workshops, project time, and what they call “guilds”—small groups focused on debate, or entrepreneurship, or engineering challenges. There are no lectures, no nightly worksheet packets, no homework in the way we usually think of it. The software adjusts itself to their skill level. If they get something wrong, it gives them another explanation and tries again from a different angle. If they get it right three times in a row, it moves them up.
For the first few weeks, I didn’t trust it. I kept looking over his shoulder at the dashboard on the parent app, which showed his “mastery” scores and weekly progress. The numbers looked good. Almost too good. My son is a bright kid but not an especially disciplined one, and I watched him sit for the full sixty-minute learning block without anyone nagging him to stay on task. That had never happened in any classroom. At home, he could not finish a seven-page homework packet without three breaks, a snack, and some negotiations. But here, he just kept going. Each small, immediate success was enough to keep him clicking forward.
What I started to notice, though, was a shift I wasn’t prepared for. When I picked him up, he didn’t tell me what he “learned” in the way I was used to. He told me what he figured out. “I got stuck on word problems again,” he said one afternoon. “But the program made me explain why I picked the answer before it would show me the next one.” He wasn’t being graded or scolded. The software just quietly refused to let him guess. That refusal, built into the system, was doing something that took years of patience in a traditional classroom to accomplish. It was waiting him out.
I remember sitting in the observation window one day during a workshop period. Six kids were building a bridge out of cardboard and tape for a contest about load-bearing design. A few of them were debating how to distribute the weight. The guide stood back and let them argue. At one point, they were all wrong in a way that was physically obvious the moment they tested the bridge. It snapped. The girl holding the ends looked at the boy who had insisted on the design, and he shrugged. Nobody said “I told you so.” They just started over with a different approach. That moment felt more valuable than any worksheet I’d ever seen come home.
But I’m a realist, and I think other parents considering this kind of school need to hear the parts that aren’t in the brochure. The first thing is that the technology works best for ten-year-olds who already have basic reading fluency and can type reasonably well. A child who struggles with decoding will spend their entire learning block just trying to get through the question, and no amount of adaptive software fixes the fact that some kids need an adult sitting six inches away, tracking their eye movement and catching the exact syllable where they stumble. The AI can’t do that. At least not yet.
The second is that the human guides are not teachers in the traditional sense. They don’t have to plan lessons or develop a curriculum, which means they spend a lot more time on emotional regulation and social dynamics. The guide we had was excellent at noticing when a kid was about to shut down. She’d pull a chair over, lower her voice, and ask a simple question: “What part feels hard right now?” She’d wait for the actual answer, then ask them to re-read the problem out loud. She wasn’t helping with math. She was helping with the moment before math, which is where most kids get stuck. In a normal school, that moment often gets missed because the teacher is managing thirty students and a pacing guide. Here, somebody sees it two minutes before the kid disengages entirely.
I’ve started using similar language at home. “What part feels hard right now?” is a question I now ask during homework, except we don’t have homework anymore, so I ask it during reading, or when he’s practicing his times tables at the kitchen table. It sounds simple, but it changes the conversation. Instead of asking “Why aren’t you working?” you’re asking “Where specifically are you blocked?” It stops the spiral before it starts.
The other thing I did, after watching how the AI handles mistakes, was to stop caring about correct answers in everyday conversation. At dinner, if he says something factually wrong about history or science, I used to correct him. Now, more often, I’ll say “Are you sure about that? Explain why.” It turns dinner into a little Socratic exercise. He’s getting used to the idea that being wrong isn’t the problem; skipping past the wrongness is where the trouble begins.
There are nights, though, when I lie awake wondering if this is going to hurt him later. The software is patient, but it’s also relentlessly positive. It never sighs, never raises an eyebrow, never communicates the mild annoyance that a human teacher might when a student is clearly not trying. Real classrooms have social pressure. That pressure is uncomfortable, but it does teach kids to perform under the judgment of others. I don’t know what happens when my son has a human professor in college who doesn’t care about his learning style. I don’t know how he’ll handle a boss who writes vague instructions and gets angry when the work isn’t done. The AI school teaches independence in a controlled environment, but the world outside is not controlled.
For now, it’s working for him. He’s ahead in math, but more importantly, he volunteers what he’s thinking about. He says “the program” like it’s a person. “The program kept making me do fractions,” he told me last week, with actual frustration. That frustration was real, and it was his own. It wasn’t imposed from the outside. It was just him, a number puzzle, and his own two hands.
We’re not committed forever. The school doesn’t pretend to be a permanent solution for every child. It’s a bridge. For some kids, it’s exactly the right shape. For others, it’s too much screen time, or too little structure, or too quiet. We’re treating it like a pair of shoes that fits right now. One foot has grown a full size already this year, and that reminds us that everything changes, including minds.
Please indicate: Thinking In Educating » We Tried an AI School Like Alpha School: What Parents Should Know