The School With No Teachers

This week, Alpha Schools announced it will open a campus in Chicago this fall. The school teaches kindergarten through eighth grade. Annual tuition is $55,000. There are no teachers.

Instead, students spend one to two hours each morning learning core subjects from AI programs on laptops. Adults in the building are called "guides," not teachers. They do not deliver instruction. After the AI sessions, students shift into workshops on public speaking, coding, and outdoor activities. The school reports that its students score in the top 1 to 2 percent on national proficiency exams.

Alpha now operates more than 35 locations across Texas, Florida, California, and other states, with over 1,000 students enrolled. The expansion to Chicago, Atlanta, and Puerto Rico is planned for this fall. The model is growing fast.

It is also raising exactly the right question, even if the answer it offers is incomplete.

What Alpha Gets Right

The instinct behind Alpha is not wrong. Traditional instruction, where one adult talks and thirty students listen, has never been an efficient way to learn. The research on this is decades old. Bloom's famous "2 sigma" study in 1984 showed that one-on-one tutoring produced learning gains two standard deviations above classroom instruction. The problem was always cost. You cannot give every child a private tutor.

AI changes that equation. A well-designed AI tutor can adapt to each student's pace, revisit gaps, and provide the kind of patient, individualized attention that a single teacher with thirty students simply cannot. Alpha is betting that this personalization, combined with human-led enrichment activities, is a better model than what most schools provide.

There is something honest about this bet. It acknowledges what many in education are reluctant to say aloud: much of what happens in traditional classrooms is not great teaching. It is crowd control, pacing to the middle, and one-size-fits-all delivery. If AI can do that part better, perhaps the human role should shift.

What Gets Lost

But here is where the model begins to thin. The scores are impressive. The question is what they measure.

Standardized tests capture what a student can recall and apply in a controlled setting. They do not capture how a student thinks through uncertainty. They do not show whether a child can hold two conflicting ideas and reason between them. They do not reveal the moment a student almost gave up on a paragraph, then found a way through. They do not record the five-minute pause before a breakthrough.

These are not soft skills. They are the hard skills of thinking itself. And they are developed not by consuming optimized content but through the friction of being questioned, challenged, and seen by someone who knows how to listen.

Northwestern researcher Liz Gerber described Alpha's approach as "self-directed learning with Montessori principles." That framing is generous and, in one sense, accurate. But Montessori classrooms are built around deeply trained educators who observe, intervene, and guide. The "guide" in a Montessori school is not a chaperone. They are a teacher by another name, one who has learned when to step in and when to step back. The distinction matters.

The Visibility Question

What Alpha's model makes visible, perhaps unintentionally, is the central tension in AI-powered education: efficiency and depth are not the same thing.

An AI system optimized for content mastery will move a student through material as quickly as possible. It will identify gaps, fill them, and advance. This is genuinely useful. But learning is not only about moving through material. It is about what happens when you get stuck. It is about the revision, the reconsideration, the moment when you realize your first answer was too simple.

The OECD's Digital Education Outlook 2026 documented this tension with uncomfortable precision. Students using general-purpose AI tools performed 48% better on tasks. When the AI was removed, they performed 17% worse than where they started. The tool carried them. It did not teach them. But purpose-built Socratic AI, the kind that asks questions instead of delivering answers, produced sustained gains that persisted after the tool was gone.

The difference is not in the technology. It is in the design philosophy. Does the AI optimize for throughput, or does it optimize for thinking?

A Different Kind of AI Classroom

At Koan, we believe AI belongs in every classroom. We also believe teachers belong there too, not as content delivery systems, but as the people who notice what no algorithm can: the student who writes a perfect essay but has stopped caring, the one whose messy draft contains a genuinely original idea, the quiet kid who just needs someone to say, "Tell me more about that."

Our AI tutor, Aidan, does not replace that human attention. It extends it. Aidan asks students Socratic questions calibrated to their rubric and their patterns. It does not generate answers. It generates thinking. And every revision, every pause, every shift in reasoning is captured in a timeline that teachers can see. Not as surveillance, but as evidence of the learning process itself.

This is what we mean by making learning visible. Not just the score. Not just the submission. The process. The struggle. The growth.

The Question Worth Asking

Alpha Schools is a $55,000-a-year experiment in what happens when you remove the teacher entirely. It will produce data. Some of it will be impressive. But the most interesting question it raises is not whether AI can replace teachers. It is what we want teachers to be for in the first place.

If teachers exist to deliver content, then yes, AI will eventually do that more efficiently. But if teachers exist to see students, to notice the thinking behind the answer, to ask the question that unlocks something no algorithm predicted, then the future is not about replacement. It is about amplification.

What if the most important thing a teacher does is not what they teach, but what they see?