Summary

Alpha School says mastery-based AI tutoring compresses academics into two hours while lifting students toward the top of national performance. Its reported SAT average is 1,410 across the entire high school, 1,535 for seniors, with freshmen targeting at least 1,410 and some already above 1,500. Joe Liemandt’s broad claim is that any student can reach the 99th percentile without spending all day at a desk or doing homework at night.
The technology thesis is not “give children ChatGPT,” but combine learning science, personalized lesson generation, and continuous behavioral measurement. TimeBack targets 80–85% answer accuracy, adapts to each student’s knowledge and interests, and uses vision models to detect guessing, skipping explanations, or leaving the learning environment. Alpha says it has invested more than $10 million in the platform, but is still spending roughly $10,000 per student on AI tokens—a major cost-down requirement.
Alpha’s operating model unbundles the five jobs traditionally forced onto one underpaid teacher. Software handles subject expertise and instructional sequencing; six-figure “guides” concentrate on motivation, emotional support, parents, and culture, with 80,000 applicants reported for roles. “Teens need high standards and high support” is the hiring and training premise, while kindergarten remains deliberately human-heavy at roughly five or six students per teacher.
Student engagement is treated as the system’s indispensable input, not a soft outcome. More than 90% reportedly say they love school, 40–60% would choose it over vacation, and two-thirds of Alpha High students once asked to remain open through summer. Joe Liemandt’s warning to schools shopping only for EdTech: unless they rebuild around the proposition that “kids must love school more than vacation, this isn’t going to work.”
The saved time becomes Alpha’s differentiated product: afternoons devoted to leadership, teamwork, entrepreneurship, financial literacy, relationships, storytelling, and grit. Fifth graders have operated food trucks and Airbnbs, while high-schoolers spend four years on evolving passion projects such as producing an all-teen Broadway musical. The model sells agency as much as academics: students should become “creators and contributors, not consumers.”
The go-to-market starts with owned private microschools, then extends the TimeBack engine across different formats and price points. New locations generally open near 25 students; Alpha retains control rather than franchising and added 13 schools in the year discussed. The group has launched Texas Sports Academy and a gifted-and-talented school, and described wilderness, Montessori, and other models for further rollout, including in 2026. Texas Sports Academy uses state vouchers to reach about $300 per month in parent payments.
The largest unresolved risks are proof, parent adoption, public-system resistance, unit economics, and consistent quality at scale. In-person charter applications in 10 states received no approvals, while Arizona authorized one virtual option; governments want substantially more evidence. Liemandt proposes a million-student, “pharmaceutical-grade” randomized trial. His ambition is 10,000 schools over 20 years and eventually one billion learners, but his blunt conclusion is that “the blocker to education reform is what’s in parents’ heads.”

Deep dive

1. The incumbent system is losing outcomes and legitimacy

Peter Diamandis opened with national proficiency at historic lows: only 35% of US high-school seniors meet reading proficiency, versus 40% in 1992; just 22% are proficient in math and 31% in science. Joe Liemandt sharpened the indictment: half of seniors graduate knowing only as much math as a 99th-percentile third grader.
Higher education no longer provides a comforting next step. The share of Americans calling college important reportedly fell from 75% in 2010 to 35%, tuition has risen 893% since 1983, and college graduates are now the group remaining unemployed longest. Peter’s conclusion was categorical: “That’s not sustainable.”
Joe’s concern extends beyond test scores. Children need enough reading, mathematics, and science to think critically, but an AI-shaped economy also demands adaptability and the ability to remain “on the frontier of what human knowledge is required.” Parents of young children are therefore asking what school should deliver 10 or 15 years from now.

2. Alpha claims elite results without an all-day academic grind

When Peter cited a 1,410 average SAT, Joe corrected the record: 1,410 covers Alpha’s entire high-school population, freshmen through seniors, while seniors average 1,535. Joe added that freshmen target a minimum 1,410, with some already above 1,500—leaving much of high school available for work beyond conventional coursework.
Their intellectual foundation is Bloom’s two-sigma finding: tutoring and mastery could move essentially every child toward the top 2%. Joe argues that roughly 10,000 subsequent learning-science papers have supported dramatically faster learning, but “it doesn’t work in a teacher-in-front-of-a-classroom model.”
Alpha distinguishes its operating evidence from its engineering aspiration. Students currently learn “over 2X” in a two-hour block, while the TimeBack learning engine is described as teaching K–12 material up to 10 times faster than classroom instruction. Joe’s broader assertion is that excellent academics need neither “all day chained to a desk” nor nightly homework.

3. Loving school is the first design constraint

Joe’s first of five dimensions is the most radical: “Kids must love school.” The remaining pillars are faster mastery, practical life skills, a redesigned adult role, and character built through culture and classmates—but he says every downstream mechanism fails if the first condition is absent.
Alpha polls students weekly and reports that more than 90% love school. When asked whether they prefer school or vacation, 40–60% choose school depending on the vacation and current workshop; before one summer, two-thirds of Alpha High students emailed asking that the campus remain open.
Joe’s challenge to inherited expectations: adults try to build companies employees love, yet accept that school “is supposed to suck” because “school’s spinach.” When conventional schools ask only for Alpha’s EdTech, he refuses the premise: the engine cannot simply be bolted onto a culture where enjoyment, agency, and saved time are secondary.

4. The school day separates focused mastery from human activity

Students begin with a “Limitless Launch,” described as Tony Robbins for children: a collective challenge that introduces growth-mindset practices. They then enter a two-hour core block, break every 30 minutes, finish academics by lunch, and spend the afternoon in collaborative workshops; there is no homework.
Personalization means two seven-year-olds can sit together while one studies algebra and the other practices multiplication. Each remains socially embedded with peers while receiving work inside the “zone of proximal development”—neither so easy that attention disappears nor so difficult that the student shuts down.
The physical environment resembles a comfortable coworking space more than rows of desks. Students may stand, share tables, or work from beanbags after earning that freedom; campuses also reserve roughly 90 minutes for unstructured outdoor play. MacKenzie Price’s point: two hours of purposeful computer use create more, not less, time for adults and children to interact.

5. Afternoons convert academic efficiency into life-skill inventory

Alpha’s afternoon curriculum covers leadership, teamwork, public speaking, storytelling, entrepreneurship, financial literacy, relationship-building, grit, and hard work. The workshops are designed as real projects rather than classroom simulations: fifth graders have run food trucks and Airbnbs, while music students form a rock band and learn branding, marketing, and production alongside instruments.
At Alpha High, students devote four years to a passion project, with permission to change direction as their interests mature. Joe’s strongest example was a senior producing an all-teen Broadway musical—sourcing performers through TikTok, securing music rights, filling a venue, and owning sales and marketing rather than merely studying theater.
The freed time permits unusually ambitious attempts, though the episode preserved the distinction between attempt and result. One student had just submitted research to Nature; Joe said no high-school student had previously been published there, but did not claim that her submission had been accepted.

6. Guides replace an impossible teacher specification

Joe decomposed a traditional teacher into five jobs: subject expert, instructional expert, student motivator, parent communicator, and administrator. “That’s a complicated spec,” yet the conventional response has been “to underpay them.” Alpha removes the first two requirements through software and demands exceptional performance on connection and motivation.
Guides do not lecture, write lesson plans, grade homework, or show children how to “carry the one.” Their role is motivational and emotional support: hold high standards, provide high support, and coach students to consult explanations, videos, and other resources until they can learn independently.
Alpha says 80,000 people applied for its guide roles, which begin at six-figure salaries. Roughly half come from traditional education—including a 2020 Florida Principal of the Year—while others come from executive work, college and professional athletics, or coaching. Traditional teachers are also the most common late-stage rejection because some struggle to stop directing and start engaging alongside children.
Kindergarten remains an acknowledged exception to automation. Alpha uses roughly a five- or six-to-one student-to-teacher ratio and retains reading specialists because its applications cannot yet reliably bootstrap four-year-olds into reading; Joe estimated another 18 months of AI progress. The discussion also emphasized that the low ratio would remain even after the technology improves.

7. Mastery changes competition from rank to remaining effort

In a time-based classroom, everyone advances whether or not they understand; Alpha students advance after mastering each level. Fewer than 10% of parents initially believe a child can score 100 on the Texas STAAR, Joe said, whereas 95% of Alpha students believe it because they have already done so.
TimeBack expresses difficulty as distance rather than identity: “You’re 17 hours away” or “You’re 5 hours away.” Joe argues that conventional performance is coded to IQ and Big Five conscientiousness—the roughly 10% who are naturally fast or willing to grind—while mastery reframes the issue as effort available to every student.
Dave Blundin’s peer-pressure concern was that ordinary schools irrationally bucket children into academic stars, athletes, or disengaged outsiders. Alpha’s answer is that every student is expected to master academics, while goals and projects provide other avenues for achievement. Peter captured the motivational reversal: school grades start at 100 and fall with mistakes; games start at zero and rise with progress.

8. The AI tutor is deliberately not a chatbot

TimeBack generates lessons from the target concept, the child’s knowledge graph, an interest graph, and cognitive-load constraints such as working-memory capacity and repetitions required for long-term retention. It aims for 80–85% accuracy: at 99% the child already knows the work, while below 66% students tend to disengage.
Asked about visual and auditory learners, Joe rejected single-channel personalization as bad learning science. Students may have preferences, but “dual coding” means presenting information visually and auditorily together produces better reinforcement than routing each child into one presumed learning style.
Peter’s screen-time pushback is worth keeping: parents are being asked to accept more computer use amid a broader backlash. Alpha’s trade is two concentrated hours of “good screen time” for four hours returned to physical, social, and creative work; even no-screen households may accept that exchange if the learning gain is real.
Chat is disabled during morning academics because, according to Joe, 90% of students use unrestricted ChatGPT to cheat: “Chatbots are cheatbots.” The rule reverses after lunch: if students use chat in the morning, they are probably cheating; if they do not use it to build a musical, company, or application in the afternoon, they are “probably failing.”

9. Continuous observation creates a learning loop

Vision models watch each learning screen and detect whether a child scrolls, guesses, skips an explanation after a wrong answer, opens ChatGPT, or plays Minecraft. The system then coaches the behavior itself—slow down, read, watch the explanation—rather than merely scoring the final response.
That monitoring is currently expensive: Joe estimated approximately $10,000 per child in AI-token costs and wants to move inference on-device. Three years earlier, humans reviewed and annotated recordings at night to classify good and bad learning, creating the examples needed to automate the process.
Alpha’s unit of measurement is XP: one XP equals one minute of focused learning. A visible “waste meter” can tell a student, “Dude, you’re wasting 50% of your time.” Motivation comes from a direct bargain—engage properly and regain the afternoon; guess or evade, and the two-hour block will not be enough.
The closed loop also exposes Alpha’s own mistakes. A new K–12 math curriculum was supposed to produce more learning in 20% less time; after eight weeks, grades four through 12 exceeded the target, but K–3 students became lost because they received too much choice. The team changed the path for the next eight-week cycle and reported getting back on track.

10. Scaffolding begins by undoing learned passivity

Salim Ismail asked whether adolescents—especially boys—need more scaffolding and whether Alpha measures stress and psychological safety. Joe said older entrants arrive conditioned to sit passively, wait for information, and move forward without mastery; guides counter that conditioning through feedback practice, growth mindset, and repeated messages that dyslexia or earlier struggles do not define capability.
Incoming assessments revealed a stark mismatch with prior transcripts. Students carrying an A in math ranged from one year ahead to three years behind; B students ranged from three to seven years behind. Joe told parents, “Your schools have been lying to you,” while promising that personalized remediation could fill the holes much faster than the years required to create them.
Every student receives a 30-minute one-to-one guide meeting each week, compared with a cited traditional-school average of 22 seconds of individual teacher attention per day. The most popular item students buy with earned Alpha currency is not a LEGO set or snack but a private lunch with their guide.
The response did not identify a formal stress metric, instead emphasizing relationships, culture, and purpose. Students complete an Ikigai values exercise and track all 168 hours of a week; guides then redirect passive consumption toward creation—vibe-code the game rather than endlessly play it, or build an audience around a useful project rather than merely scroll TikTok.

11. Private microschools are the initial distribution wedge

Alpha normally opens a city with roughly 25 students and grows from there. Joe’s adoption curve is blunt: “Nobody wants to be the first 20,” but once a campus has 100 students, many more families become comfortable. Peter said a prospective parent needs both a recommendation from a trusted local adult and evidence of an enrolled child doing something their own child cannot.
Peter recounted MacKenzie’s founding-family story: when she was co-founding Alpha, she wanted to find the 20 people who would surround her children. Joe admitted he remained a skeptical “laggard” for two years before moving his daughters.
Alpha owns and operates every location rather than franchising because the founders want direct quality control. The organization added 13 schools in the year discussed, uses shadow days to demonstrate the model, and warns parents not to send children unless they are genuinely prepared to transfer: many refuse to return to their old school afterward.
Joe frames Alpha as a high-end, “price is no object” private school inside a roughly $50 billion market. Peter cited broader public K–12 spending above $1 trillion—more than $20,000 per child and above $40,000 in some cities—and argued that this should be sufficient to provide both learning and a day children enjoy.

12. TimeBack enables new schools, but public systems remain resistant

The two-hour engine permits multiple afternoon identities. The group has launched Texas Sports Academy and a gifted program and described wilderness and Montessori schools, among other models. Joe said additional models based on TimeBack would roll out in 2026. The gifted version targets children who answer “more academics” when asked what would make them love school; the sports version begins serious training at noon rather than after 4 p.m.
Texas Sports Academy uses part of the state’s newly allocated billion-dollar voucher program and targets roughly $300 per month in parent payments. This is the founders’ route toward lower price points: preserve TimeBack’s academic engine while varying the afternoon experience, staffing, and facilities.
Public adoption is a later ambition. MacKenzie applied for in-person charters across 10 states and received no physical approvals; Arizona did authorize a virtual option. A school-board member reportedly declared, “I was put on this earth to stop people like you.” For now, they expect private distribution to occupy the first decade and public systems perhaps the second.

13. Scale depends on proof, cost-down, and relentless adaptation

Peter’s closing pushback was direct: “Is everything perfect? What do you guys worry about?” Joe named parent disbelief, maintaining consistent outcomes across every new guide and campus, defining testable life skills, and reducing AI cost. Even grit needs an operational proxy; Alpha currently uses accomplishments such as every third grader solving a Rubik’s Cube.
The stated ambition is 10,000 schools over 20 years, yet Joe’s real target is one billion children. That requires motivating students even when Alpha cannot return four hours of their school day, so the company is working with video-game designers and major influencers on motivation-plus-learning products intended for release in 2026.
Governments’ caution received an unusual concession: education often deploys interventions to 55 million children without rigorous trials, while bad learning science persists. Joe wants a million-student randomized controlled trial—“pharmaceutical-grade testing”—so officials and parents can determine whether the model works rather than dismissing it as a hoax.
The founders also reject a fixed curriculum for a 12-year technological horizon. Alpha’s “Brainlift” separates knowledge LLMs handle from what humans should retain, but Joe conceded that Grok 7 might invalidate today’s boundary. The school updates on six-week cycles; since its 2014 founding, the decisive change was generative AI making billion-child scale imaginable—and the lesson Joe wishes he had learned sooner is that parents, not students, are the reform bottleneck.

Universal Medical Intelligence: OpenAI's Plan to Elevate Human Health, with Karan Singhal 🔬Searching the Space of All Possible Materials — Prof. Max Welling, CuspAI