Two Academic Hours a Day, and 96 Percent of Kids Say They Love School

Sarah Guo and Elad Gil introduce Joe Liemandt as the founder of Trilogy, a 1990s software company he calls the first product of that decade to sell a billion dollars of AI, and as the principal of Alpha School. In high school he wrote a paper on AI with a paragraph that said neural nets were decades away. At Stanford he sat in a class with Edward Feigenbaum, the father of expert systems, then dropped out and spent 25 years on classic SaaS: sales configuration, organic builds, and acquisitions. You could not even call it AI then, he jokes, because AI was a bad word.

The school arrived through McKenzie Price, a Stanford graduate he had hired into Trilogy in the 1990s. A decade ago she started Alpha, a very different campus that used apps in the morning and no traditional teacher at the front of the room. Liemandt spent two years telling her he would not send his kids to her weird school. Every parent, he says, wants children educated the way they were. He had gone to Catholic school, had not even liked it, and still defaulted to the local version of the same two-hundred-year-old model.

The hang-up was specific. Price was using DreamBox; he could not believe an app with no teacher would beat a human at math. Good school equals good teacher, and everybody knows it. Opening a new Alpha location still hits the same wall: once a hundred kids are in the building, families want in; nobody wants to be among the first twenty. Price wanted to be a founding family partly so she could choose the other kids around hers. It took two years. The daughters are now best friends and have been for a decade.

Three years ago, when generative AI arrived, he decided neural nets were finally here and the magic of Alpha could leave Austin. Point solutions in education had always existed; they did not scale because great teachers do not multiply. He is a product person, so he made himself principal: what happens when fifth graders fight, what do parents yell about, and if families drop children at a building with other kids and other adults, how would you unleash human potential with 12 years to reimagine it? The goal he states on the show is a billion kids.

Ten years ago Liemandt asked Price's co-founder what he did not know about school. Two things, the man said. First, kids must love school. Second, once they do, your expectations of them are too low. Call me back in a hundred days. Both turned out to be true. With generative AI in hand, Alpha's first design rule became: kids must love school more than vacation. They survey students. Ninety-six percent say they love school. Every eight weeks, depending on the vacation and how afternoon workshops went, 40 to 60 percent say they love school more than vacation. If you are going to hold children in a building for 12 years, he thinks every parent should expect that.

In his first week as principal he asked incoming fifth graders whether they loved school. The whole class said no. What would make them love it? Less school. How much less? None. School, the hosts note, is a bundle: academics, physical activity, socialization, and childcare. The kids meant the academic part. Six hours in class would never produce love. They negotiated two hours of genuine engagement in the apps in exchange for four hours of "awesomeness." The students said yes. He went back to his team: jam everything into two hours, and it has to stay in their heads.

Edtech has been deployed for years while test scores keep falling, he says, because of engagement. Even a decade-old DreamBox could work; studies suggest only 5 to 10 percent of people actually engage with Khan Academy, the self-motivated slice. Learning-science papers going back 40 years already claimed kids could learn two, five, or ten times faster. Those methods do not work with a teacher in front of a classroom. The team pulled methods out of a book by Dan Schwartz, dean of Stanford's graduate school of education, chapter by chapter, and built a learning engine on those principles. The commitment to kids is now top-one-percent academics in two hours a day. When the app goes green, the afternoon opens for life skills: leadership, teamwork, grit, entrepreneurship, financial literacy, storytelling, public speaking, relationships. That is the bundle parents actually care about once AI "knows everything."

A fifth grader's day starts with Limitless Launch, which Liemandt describes as Tony Robbins for kids: growth mindset, the claim that children are limitless if the environment lets them be. Then they sit with an AI tutor for two hours of personalized lessons. Age grade and knowledge grade, he insists, are two totally different things. In the average American sixth-grade class, very few students actually need sixth-grade content. A sixth-grade teacher's job is still to deliver it. That mismatch, in his view, is why scores keep falling.

Hundreds of students had just transferred in as campuses opened. His first academic talk to parents is blunt. Your $50,000 private school has been lying to you. On Alpha's standardized tests, a straight-A transcript typically lands somewhere between one grade level ahead and three behind. A B student lands three to seven years behind. He has incoming freshmen, from high-end private schools, who cannot write a grammatically correct third-grade sentence. The consolation is speed. The average grade-and-subject combination takes 20 to 30 hours to master. Fourth-grade math on Math Academy takes 28 hours; fifth-grade 26; sixth and seventh 22 each. Parents imagine hundreds of hours of catch-up. Sixty hours behind is 60 days if the child adds a third hour at school or as homework. Those are hard, focused hours, not the drooling-and-doodling elementary afternoons he remembers.

The engine's job is to keep each child in the zone of proximal development, at 80 to 85 percent correct. Ninety-nine percent means you are serving content they already know. Drop to 66 percent or below and they disengage. Every video-game designer knows that curve. A classroom teacher cannot generate an unending stream of items in that band for every student; an AI tutor can, including scaffolding lessons when the leap is too big. Learning science, he says, does not support throwing kids way over their heads. If you handed him an MCAT and told him to show grit, he would fail; roll him back to freshman biology and build up. High-schoolers who cannot write go through third- and fourth-grade lessons in hours, not years, and in weeks can be producing AP Language free-response work. Perfect content, not productive drowning, is the acceleration.

Afternoons are project-based workshops: something the child will love, plus a life skill. Love of school, Liemandt says, requires three beliefs he would wave a wand to install in every parent. Kids must love school as much as vacation. They can crush academics in two hours a day. And the key to a child's happiness is high standards. Kids want to do hard, awesome things and to accomplish them. Low standards steal the love. Hosts note that apprenticeship across centuries already worked that way. The catch is that parents often reject the claim for academics even while they fully believe it for sports.

Everybody wants their kid on the championship team with the coach who says this will be the hardest thing we have ever done. Last place last year, first next year, grind, teamwork, leadership. That is why families buy after-school sports: to purchase those life skills. Alpha tries to run the same ethos all afternoon. Kindergarteners climb 40-foot rock walls. Parents, whose job is safety, do not want to watch a child struggle, fail, and sometimes cry on the way to success. Every child-development expert, he says, wants that cycle run as often as possible, supported by a caring adult. It is also, for Alpha, the key to loving school.

Alpha does not hire adults to teach seventh-grade science. It hires motivational and emotional-support specialists, and the profile changes with age: kindergarten is wrap-the-kid-in-love; high school includes former NFL athletes and NBA coaches who can take video-game-playing middle-school boys and say we are going to go do something. Guides start at $100,000. The future of teaching, in his telling, is the reason people became teachers in the first place. Your best teacher found you and convinced you that you could do awesome stuff. Nobody became a teacher to grade the science quiz, and nobody remembers them for markup.

Morning academics run in 25-minute pomodoros of math, science, language, and reading. During a block, a guide will pull a child out: skip math today, tell me about the weekend, the softball tournament, why engagement is off. That is the full-time job. At the San Francisco campus, 23 days in, a parent told him the guide already knew more than teachers had in years. High schoolers interview prospective guides before they are hired, a practice outsiders assume will tank standards. Kids who want to be great, he says, pick the strict coach. Steph, who runs K through 8 and came out of sports, defines her job as bringing the athletic ethos to academics.

Guides are accountable for three commitments to every child in the group: love of school, 2x learning, and life skills. Miss them and you will not be at Alpha long; it is not a tenure system. Students are surveyed on whether the guide is one of those one or two adults who transform a life. For middle and high school, parents are asked a sharper question: do you trust the Alpha guide to hold high standards so you can provide unconditional love? Liemandt's oldest daughter would not let him read her Stanford application. Chloe, her Harvard-graduate guide, told him not to worry. It is the same outsourcing parents already accept from a sports coach. Lead guides can earn more; the relationship is how he thinks teachers become 10x.

The billion-kid goal means designing for personality as well as price. Besides Alpha, they have opened branded schools with different afternoons: a sports academy, a gifted-and-talented campus doing robotics, math olympiad, and book-writing. When only 40 percent of kids prefer school to vacation, guides are asked how to juice it to 75. GT guides want more academics, a third power hour. His oldest daughter read calculus books in high school, is a double major at Stanford, and wrote an equation-based English class because the STEM boys cannot write. His younger daughter, still in high school, wakes up asking whether this is the last academic work she has to do. She has millions of TikTok followers and built an AI dating app for high-school boys. The continuum is the point: find the passion, give them more of it in the afternoon, and morning academics ride along.

The product is named Timeback. The single biggest motivator, he says, is don't waste my life, don't waste my 12 years. Traditional class is 95 percent wasted time, like sitting in traffic. An Astral Codex Ten post by a GT parent ranked Alpha bucks as the top incentive. Liemandt corrects the record. You cannot pay Westlake High kids enough to enjoy six hours of class plus four hours of AP homework. Tell the same students they can still post a 1550-plus SAT and fives on APs in two to three hours a day, plus four years of afternoons they love, and the bargain works. Stickers, leaderboards, and competition still matter; every child is different. Guides invent programs daily under a high-support, high-standards brief.

Money is the controversial slice, and he does not dodge it. Kindergarteners earn Alpha bucks in a miniature economy of earning, saving, spending, and donating. In middle school, students who did not enter in the top one percent get $1,000 when they hit it in a subject, deposited into a Robinhood for kids account as real money to invest. Some YOLO it and lose it; parents yell. Fake shadow accounts, he argues, teach dysfunctional gambling because there is no downside. One girl used the money to fly to New York, meet producers, and push an all-teen Broadway musical; her math SAT reached 790. Adult achievement already works this way.

The popular claim that extrinsic rewards wreck intrinsic motivation is, he says, just not true. Kids carry inherent blocks: I'm not smart enough, I'm not a top-one-percent kid, I don't have the habits. If money builds daily habits or breaks a block, every parent should pay it. His younger daughter arrived in middle school content with the top 10 percent; the older sister was "the smart one." He bet her $1,000, a gift card she folded up and looked at every day, that she could not hit the top one percent. After she did, she sat with her mother, who had hated the experiment, and said she had never thought she could do it. She now believed she was as smart as her sister if she put in the work. America spends about $20,000 a year per child. A thousand dollars to remove that story, he thinks, is cheap.

When the Secretary of Education visited Alpha the week before the taping, Liemandt offered one national fix that uses money. The median American high-schooler gains one point on a 300-point scale over four years. The top 99 percent still rise about six points a year. Learning slows in middle school and nearly stops for the median student in high school because the system is time-based, not mastery-based. Every learning-science paper, he says, describes Swiss-cheese holes. If you have not mastered fractions, algebra is hard; if you have not mastered algebra, chemistry goes next. Knowledge is hierarchical.

Three years ago he sat seventh graders down and told them to fill a fourth-grade hole. They did not want to go backward. Their parents wanted it even less, and yelled: my kid needs to be in eighth grade. He gave up on the parents and went to the kids. Score 100 on the Texas STAAR and he would hand them a hundred-dollar bill, any grade level. STAAR starts in third grade. They cleaned up third and fourth, then hit fifth and scored an 85 or 75. The AI tutor generated the missing lessons. They looked at the list, said they could do it in a week, and moved up. The program does two jobs: it fills the holes so learning can stay linear, and it breaks the story that only gifted kids get 100 on a state test.

In ordinary school, he argues, being "good" is IQ-coded and conscientiousness-coded. Alpha's system tries to separate those. Personalized lessons at 80 to 85 percent, plus time, and you can master every subject through eighth grade. The athletic ethos never crossed into academics, in this telling, because the time-based factory convinced parents that school is an intelligence contest. A $400 or $500 incentive per child, against roughly $20,000 of annual spend, is the program he wants run for everybody in America.

Gil puts Alpha in the wider weather: *The Coddling of the American Mind*, autism diagnoses moving from one in a few thousand to 3 percent of the population, mass ADHD and Ritalin, a teenage-girl mental-health crisis and SSRIs. Liemandt does not deny the numbers. He thinks a share of them is what you should expect from a 12-year system of low expectations, 95 percent wasted time, and endless pressure. An IEP, in his reduction, is a parent fighting a public school for the personalized education everyone already knows a tutor beats. Families arrive with dyslexia, dysgraphia, ADHD labels. At Alpha the child is not sitting bored for six hours, is not the "problem child" who has to be medicated to survive the room. Two hours of engagement and an afternoon of something else, he says, completely transforms many of those kids.

Middle-school girls show up saying they are not math-science girls and then scroll TikTok all day. Workshops start with a values chart and a vision board, then ikigai, then a 168-hour audit of the week. One girl told him she guessed she was going to be the best TikTok scroller, the opposite of her vision board. The school's job is to turn consumers into creators. His claim is that kids do not wake up wanting to be video-game-and-TikTok professionals; ordinary school just gives them no other exit. His younger daughter's millions of followers and the dating app sit on the creator side of that line.

Alpha itself is the expensive brand, designed as if price were no object, planned for a hundred cities. Beneath it they are building cheaper schools, now down to $15,000 a year, below what the average public school spends. A sports academy needs a field and a coach, not a stack of boutique workshops, and can run 25 to one. In ordinary school, smaller classes imitate tutoring. Alpha's high-school guides, offered double pay for double the group, all voted for the larger group and a more awesome coach. Gifted campuses at $20,000 to $25,000 do math olympiad and robotics. A next-gen Montessori sits at the same band. They expect hundreds of Texas sports academies. Texas spends about $100 billion a year on education and had just released a billion dollars in vouchers, roughly $10,000 to $11,000 per child, so a $15,000 tuition is four or five hundred dollars a month from parents.

The company is for-profit, with a large scholarship fund beside it. Liemandt wants to kill the idea that bringing capitalism into education is bringing evil into the world. He is recruiting builders: Timeback will open as a platform with the learning science packaged, so someone else designs only the afternoon. Texas sports academy is coaches in a thousand districts. Most edtech, he says, does not use learning science; rebuild from those papers and kids can learn a subject in 20 hours. A triple-A game studio is building on Timeback, free to the planet and, the team believes, potentially the most profitable game ever, because 10 million American moms would pay $100 a month to put a child in the top one to five percent. He wants the full stack: engineers, operators, buildings, guides.

Guo asks about the wall of public-school regulation. In year one he put in the first billion as a seed fund to find product-market fit. Rebuilding education, he says, will take hundreds of billions of dollars and 10,000 buildings. Every country will want its own model; education is the sovereign-AI use case governments will actually talk about. You do not have to fix public school on day one. The U.S. private-school market is already over $50 billion and has not innovated in a hundred years. A million American kids can afford $50,000 tuition. Vouchers are opening underneath: Texas's billion, and a federal bill Bloomberg had just scored at about $200 billion flowing to private education. The reason so little has changed is that 99.9 percent of parents cannot imagine school without a teacher at the front of the room. Price needed two years to move one family. Break that image, he says, and everything can be rebuilt.

Public school, he thinks, is a decade away. Education is fad-prone; his own n is not big enough. He wants a million kids and pharmaceutical-grade randomized trials, which the field almost never runs. Recent adoptions he calls terrible include dropping memorization of multiplication tables. Cognitive-load theory, in his version, says working-memory slots fill up; if 7 times 8 is still a calculation in the middle of an advanced problem, you are doomed. Kids transferring in, public and private, do not know those tables because it is now considered bad. A bigger wave is already hitting: ChatGPT. Give it to kids for academics and 90 percent will use it to cheat. "It is a cheat bot. It is not a chat bot." Alpha's two academic hours have no chat functionality. They tried; students jailbreak it.

Writing, he argues, is thinking, the way you drive knowledge into long-term schema. Kids write one prompt and think they are writers. A parallel fashion says facts need not live in the brain because Google, then ChatGPT, can fetch them. The same parents then say they want critical thinking. No learning science supports reasoning without a fact base. When an LLM tries it, we call it hallucination; kids doing the same are just making things up. AI can still giveth: load facts the way Neo uploads skills in *The Matrix*, ten times faster than a six-hour day. Policy, if he had the room, would be: do not deploy ChatGPT in its current form, and rebuild products so the model is informed by learning science.

Public-school leaders have a harder product job than he does because they must serve everyone. In Austin, a superintendent tried to close long-failing campuses and move kids to a stronger academic school; parents protested because an older daughter had a teacher who transformed her life, and they would trade seventh-grade math for that adult. The bundle fights itself. School choice is one answer. Alpha's public pilots cover a few thousand students, often MTSS tier three, the bottom 10 percent, where schools have the most flexibility. In Fremont the hard part was not whether the app worked. Motivation was 90 percent of the answer. Kids still had to sit six hours; why engage? Gift cards finally moved them. Edtech dropped into an unmotivated classroom does not matter, AI tutor included. Kids cheat with AI because they are trying to get through the system, not to learn. Five to 10 percent are self-driven and will love a Socratic tutor. The other 90 percent, asked for the highlight of their day, say lunch, friends, recess.

Edtech failed as an investment, he says, because it sells to a system that does not actually optimize academic outcomes, has no ARPU, and runs endless sales cycles, and because the products are not that good. Rebuild from learning science. Math Academy's results are a step function: 28 hours for fourth grade, 26 for fifth, 22 for sixth and seventh, backed by a 500-page book. Other math apps do not do that. Meanwhile, stop selling to school boards. Parents will spend about 10 percent of AGI on education their kids love; societies spend about 5 percent; it is a $7 trillion industry. The payment path that already works is parents. The number-one national program he would scale tomorrow is still 100-for-100.

Guo asks why a financially independent founder, with children already through or almost through both the old system and the new one, would walk back into this mess. Because it is awesome, he says. The prior career was fine; there was nothing to complain about. The last three years as principal have been so much better. Nothing is more important for a society than raising the next generation; it is the definition of whether the society continues. Kids are awesome. Transform the first 12 years and the rewards are 10x, student stories and parent stories he did not even have time to tell.

His own kids are moving on after 20 years with them. The question is what to do with the next 20. He thinks these will be the best 20 years of his life. The closer to builders is that the work is societally important and that new insights, including business and capitalism, have to come into education. He says he could not look forward to anything more.