Joe Liemandt: Alpha School and the Future of Education

Liemandt opens with a stark diagnosis: the education system Americans grew up with is fundamentally broken and will not prepare children for an AI-driven world. Even before AI became a household concern, academic results were declining year after year, despite increasing investment. Parents feel the tension acutely — they know the old model isn't working, but they don't know what should replace it.

The root cause, Liemandt argues, is structural. Traditional schools are time-based systems where students advance by age, not mastery. Success is coded for high IQ and high conscientiousness — if you're a natural grinder with strong cognitive ability, you thrive; if not, you fall behind. The system doesn't adapt. Pouring more money into a time-based, teacher-in-front-of-classroom model doesn't raise IQ or conscientiousness, so outcomes stagnate. Meanwhile, the disconnect between educational credentials and earning potential in the U.S. has led parents to deprioritize academics in favor of vague «life skills,» creating a culture where 80% of Harvard students receive A's and grade inflation masks the collapse of standards.

Liemandt distinguishes sharply between chatbots (which students use to cheat) and generative AI tutoring systems that create personalized lessons. The Alpha engine, called Time Back, uses an LLM fed with four inputs: the curriculum to be taught, the student's knowledge graph (what they know and don't know), their interest graph (what they care about), and cognitive load theory (how their working memory functions). The AI generates the perfect next lesson — a worked example in math, a passage at the right reading level, a video explanation tailored to gaps in prior knowledge — and adjusts difficulty in real time to keep students in the «zone of proximal development» where questions are 80–85% correct.

Crucially, the system enforces mastery: students cannot advance until they demonstrate 90% proficiency on a topic. This breaks the IQ barrier. In traditional schools, a student who gets 80% on a test moves on; at Alpha, gaps are filled immediately. As a result, learning becomes effort-based, not ability-based. Every child can achieve 100% on state tests if they put in the hours — and the system tells them exactly how many hours remain. Sports analogies resonate: no basketball coach would let a point guard lose the ball 20% of the time, yet schools routinely pass students who miss 20% of the material.

At Alpha, adults are called guides, not teachers, because they don't deliver academic instruction — the AI does that. Instead, guides focus entirely on what great teachers have always done best: connecting with kids, motivating them, and holding them to high standards while providing high support. The job spec narrows dramatically. Guides don't need to be domain experts in seventh-grade science or know the latest learning science research. They need to be the world's best at building relationships with children and coaching them through struggle.

This redesign solves the teacher shortage. Traditional teaching requires five hard-to-find skills: domain expertise, pedagogical knowledge, classroom management, parent relations, and administrative compliance. Alpha eliminates the first two via AI, assigns parent relations to a dean, and focuses guides purely on mentorship. The result: 50% of educators drop out when told they're accountable for every child's learning, but those who stay are measured on whether students achieve the three commitments. Guides conduct one-on-one check-ins during the two-hour academic block, asking about weekend plans, goal progress, and engagement barriers. They hold students to standards no parent could enforce alone — especially in adolescence, when teens resist parental judgment.

Liemandt struggled to convince seventh graders to fill knowledge gaps from fourth and fifth grade — they saw it as stigmatizing. His breakthrough was the 100-for-100 program: students pick any grade level, take the state test, and earn $100 for a perfect score. A seventh grader might start with third grade, ace it, move to fourth, stumble at 75% on fifth, and then have the AI tutor generate the missing lessons. Within a week, they ace fifth grade too. The experience rewrites their self-concept: «Of course I can get 100 on any test. It's just a matter of how many hours.»

This mindset shift is profound. In traditional schools, fewer than 10% of parents believe their child can score 100% on standardized tests. At Alpha, every child believes it — because they've done it. The program also reveals a systemic lie: high-end private schools charging $40,000–$75,000 give out A's to students who are three to seven years behind grade level. Parents only discover this when they take Alpha's diagnostic. Liemandt describes meetings where parents learn their straight-A child can't answer fourth-grade questions. The schools avoid AI tutoring because it would expose the gap between inflated grades and actual knowledge.