A Billion-Dollar Education Bet Says Children Can Learn Faster With AI

Joe Liemandt describes Alpha School as a 20-year attempt to “fix education” at global scale. The premise is not that school should be marginally improved, but that the basic allocation of a child’s day should change: two hours for academic learning, the rest for the skills and experiences he thinks conventional schooling leaves underdeveloped.

His claim is that AI and learning science now make it possible for children to learn far faster than the classroom model assumes. Alpha’s promise, as he frames it, has three parts: students should love school, learn twice as much in two hours a day as they would in six hours of class plus homework, and spend the remaining day on “valuable life skills” such as leadership, teamwork, entrepreneurship, financial literacy, storytelling, public speaking, relationship building, socialization, grit, and hard work.

The central tension is not only whether Alpha’s technology works as Liemandt says it does. It is whether parents and schools will accept a radically different bargain: fewer hours of conventional instruction, fewer adults lecturing academics from the front of a classroom, more AI-mediated learning, and more deliberate exposure to hard tasks that look age-inappropriate to many adults until children complete them.

The model is intentionally demanding. Shaan Puri opens with the line that “kindergarteners must climb a 40-foot rock wall and pass a receive-critical-feedback-without-crying test.” Liemandt’s response is not defensive: “This is the best time in history to be a five-year-old.”

That sentence contains the center of his argument. He believes parents underestimate children’s capacity because they are anchored to the schools they experienced. In his telling, Alpha is not asking less of children by using AI to reduce academic seat time; it is asking more of them by freeing time for hard, concrete challenges.

Alpha’s examples are deliberately physical and visible. Kindergarteners climb a 40-foot rock wall. Second graders run a 5K. Eighth graders complete a Tough Mudder and must cross the finish line together. Students launch businesses and stage Broadway-style musicals. Liemandt says the most motivating activities are those where children can beat their parents “fair,” not because the parent lets them win. For him, those moments are evidence that children’s perceived limits are artificially low.

He is explicit that the hardness is not incidental. It is the product. “Kids want to go do hard things,” he says, but only if adults provide both standards and support. A child should not be told simply to climb the wall and left alone. The adult role is to scaffold the challenge so the child can do something that initially looked impossible.

Liemandt links the education bet to the operating style he developed in business. At Trilogy, the software company he founded after dropping out of Stanford, he built an organization around high difficulty, elite recruiting, and extreme standards. Alpha, in his account, uses related ingredients — a hard problem, a high bar, a recruiting challenge, a simple product message, and a belief that people are most alive when working on something difficult enough to matter — but applies them to children and schools.

The public arc matters because Liemandt had been largely absent from the limelight for two decades. A 1996 Forbes cover shown in the piece features Liemandt, then 27, under the headline “They Keep Getting Younger,” with the subhead that “Joe Liemandt, 27, and the kids at Trilogy software are out to change the way people buy and sell.” Another on-screen article headline, attributed to Newsfinance and dated October 17, 2025, reads: “Joe Liemandt Invests 1billion Into Education Startup.” A third visual, from The Week US, frames Alpha as a public controversy: “Alpha School replaces teachers with AI. Is the future of education here?” Its visible subhead says the Department of Education is championing the model, “but critics are not so sure.” The images establish the contrast: a 1990s software founder reappearing with an education bet he says has already absorbed a billion dollars of his own money, into a debate where the reception is not simply credulous.

Joe Liemandt’s first company began with a technical idea from expert systems: configuration. He had written about AI in high school, including a line that neural nets were “decades away.” At Stanford, he says, he studied with Professor Feigenbaum, whom he calls “the father of expert systems at AI.” He and his friends decided they could solve a configuration problem that companies had been trying to solve since the late 1970s.

The product was a configurator. Liemandt explains it by starting with something familiar — a car or Dell computer website where a buyer selects options — and then scaling the complexity up “a million times” to room-sized phone switches, Boeing airplanes, mainframe computers, and large manufacturing systems. Sales representatives were selling complex products in the field, only for the manufacturer to discover later that the configuration could not be built. Trilogy’s product put the knowledge of a valid configuration into the hands of thousands of sales reps, so they could sell products that manufacturing could actually deliver.

The economics were the point. Liemandt says the problem was costing Fortune 500 manufacturers “zillions of dollars,” and that if Trilogy could solve it, customers would pay “millions and millions of dollars.” He also says Trilogy’s 1990s product was “the first AI product to sell a billion dollars.”

That success did not come quickly. He says it took three and a half years to build the product and persuade anyone to buy it. Fortune 500 companies did not want to buy from college dropouts. They would try Andersen Consulting, Oracle, or other established firms first. But if those firms could not solve the problem, they came back.

Liemandt’s pricing philosophy followed from that position. Trilogy was, he says, “the most expensive software you could buy in the 90s,” because the company understood that it was the customer’s last choice — and that the product was worth hundreds of millions if it worked.

This is also how he connects his upbringing to his later operating style. His father had been Jack Welch’s strategic planner at GE, and Liemandt says he absorbed GE’s language of gross margins, ROI, and value selling early. When he told his father that televisions were the only cool product GE made, his father answered by comparing the gross margin of a CAT scanner with a commodity TV. Liemandt says that background helped him understand that he was not selling software features; he was selling economic value.

His education thesis borrows from that same product-management and go-to-market instinct. He describes Alpha as the combination of product management and business discipline applied to schooling: if you understood the user, the value proposition, the economics, and the operating model, how would you rebuild education?

Joe Liemandt’s account of recruiting is one of the clearest windows into his management philosophy. He says Trilogy hired 2,000 Ivy League, MIT, and Stanford graduates and brought them to Austin at a time when the city did not have a large software engineering base. In his telling, the company’s reputation in the 1990s was that it was the only firm able to consistently beat Microsoft for top college hires.

He says Microsoft won 90% of college graduates when competing broadly, but Trilogy won 70% head-to-head. Bill Gates, he says, flew to Austin to understand why. At dinner, Gates went person by person through candidates who had offers from both companies, discussing salary and technical capability in detail. Gates then said he would personally call every candidate with competing offers.

Liemandt’s response was escalation. He says he told Gates, in effect, that Trilogy needed the candidates more. If Gates called them, Trilogy would bring them on a ski trip and ask how it felt to receive the last call they would ever get from Bill Gates, because Microsoft had tens of thousands of employees and Gates would never have time for them again. Liemandt says Gates, Steve Ballmer, or Mike Maples did call Trilogy candidates for a period, and Trilogy kept escalating.

The deeper recruiting insight, as Liemandt frames it, was that ambitious young people did not want the easiest path. Sam Parr and Shaan Puri contrast this with the Silicon Valley perk model: laundry, haircuts, free food, convenience. Liemandt says Trilogy University was “the hardest hundred days of their life,” and that difficulty helped the company recruit against larger, safer employers.

A Fortune 10 board on which he served once asked why mechanical engineers would choose Trilogy over the larger company. Liemandt says his answer was that the Fortune 10 internship had been made easy, so the students were not doing anything significant. At Trilogy, they would do the hardest thing they had ever done.

The same logic explains some of Trilogy’s most infamous rituals. Parr brings up a story that Liemandt took recruits to a casino and gave them money to gamble, using their willingness to risk it as a fit test. Liemandt corrects the detail: it was their own money. At the end of Trilogy University, if enough teams had hit their goals, the company would take everyone to Las Vegas, gather around a roulette wheel, and require each person to bet one month’s salary on a number. Some refused. Liemandt says the point was to confront people who claimed they wanted to be entrepreneurs and risk-takers but would not take what he considered a small risk.

He says The Wall Street Journal wrote a cover story about the practice and that parents gave him “a lot of flak.” But he presents it as consistent with the culture: pressure, risk, and a demand that people reveal whether their self-description matched their behavior.

Trilogy University also had what Liemandt calls a “swap program” with the Navy SEALs. SEALs came into Trilogy’s program, and Trilogy employees went to California to experience part of SEAL training. He says the SEALs considered Trilogy’s mental standard and rigor comparable to their physical standard. Parr jokes about whether it is easier to make “a meathead good at programming or a nerd good at swimming in cold water.” Liemandt answers that SEALs are “super smart” and heavily filtered; the point was not intelligence but operating under a different kind of decision environment.

For Liemandt, the pattern starts in kindergarten and extends through late career. People want to be with other capable people working hard on something difficult. He cites startups, SpaceX, and the Navy SEALs as variations on the same motivational structure: camaraderie around a hard mission.

Joe Liemandt distinguishes his preferred model from the caricature of the impossible boss. Shaan Puri raises the “Steve Jobs asshole” version of high standards, where nothing is ever good enough. Liemandt says that is not the model. The model is “high standards, high support.”

He says most parents, managers, and institutions fall into one of two broken patterns. High standards with low support means throwing someone into a hard problem and telling them to figure it out. They may struggle for a while, but many eventually disengage. High support with low standards avoids stress and protects the person from hard challenges, but it also prevents resilience, grit, and self-confidence from forming.

The alternative is scaffolding: keep the standard high, but make the path visible and achievable. His example is Alpha students taking the Texas STAAR test. He tells seventh graders they can all get 100s. They tell him it is impossible. He offers “100 for 100s” — $100 for a perfect score — but with a catch: they can take any grade-level test.

A student might start with a third-grade test, get a 100, then try fourth grade and get another 100. On fifth grade, the student might get an 82 because prior schooling did not produce mastery. At that point, Liemandt says, the AI tutor can identify missed concepts and provide lessons to close the gaps. The work might take a week. The student, now motivated by the next $100 and by the experience of prior mastery, does it. By the time the student reaches seventh-grade material, the mindset has changed. The task is no longer impossible; it is a sequence of work.

The management lesson is the same. Telling a company to “go build a billion-dollar product” is only a high standard. The manager’s job is to provide the equivalent of grade-level progression: the training, examples, tools, and motivation that let people see how to get there. Liemandt applies this critique to companies trying to become “AI-first.” Handing employees Claude Code and saying “good luck” is, in his view, high standards without scaffolding. Leaders have to model what excellence looks like and show people how to repeat it.

Liemandt credits two people with shaping this view. At Trilogy, his head of HR, Jim Abel, coached him out of what Liemandt calls a high-standards, low-support default. Abel helped him understand what organization he needed if he wanted to bring a 2,000-person company with him. On the education side, he credits psychologist Dr. Yeager and recommends Yeager’s book 10 to 25: The Science of Motivating Young People. Alpha’s adult role, he says, is based on a “mentor mindset”: high standards, high support.

Liemandt cites a specific feedback pattern from Yeager’s work: telling a student that the feedback will be hard, but that it is being given because the adult believes the student can succeed. The support is not indulgence. It is a signal of belief paired with concrete coaching.

Joe Liemandt says he once wrote 20-page strategy documents. Abel told him they were useless. Liemandt initially rejected that criticism, arguing that Trilogy employed Stanford and Harvard graduates who could read a sophisticated document. Abel’s counterargument was organizational rather than intellectual: a long strategy lets every employee find one sentence that justifies what they already want to do and declare themselves aligned.

To prove the point, Abel interviewed every person in the company and then presented what employees thought Trilogy’s strategy was. Liemandt says the result was “the opposite of what we’re doing.” That convinced him that strategy had to be compressed. Abel’s rule was “three lines, three words each.”

Liemandt still resists calling that sloganeering. The test, he says, is whether the words have edge. If no one could plausibly say the opposite, the phrase is wasted. “Integrity” fails because no company positions itself as the non-integrity company. “Love school” passes because many people do believe school need not be loved; they see enduring school as part of learning grit.

Alpha’s three commitments are his current version of that compressed strategy.

He says 96% of students said they loved school when Alpha asked the simpler question. Eighteen months before the interview, Alpha raised the bar by asking whether students would rather go to school or go on vacation. Four weeks before the interview, he says, 46% of students chose school over vacation.

That metric matters to him because it rejects the idea that rigor and enjoyment are opposites. He says he cannot build a school kids love more than vacation with low standards. The excitement comes from doing hard things with friends and getting better.

The second commitment, learning twice as much in two hours, is where Alpha’s AI claims sit. Liemandt says children currently spend six hours a day in school plus homework. He says Alpha’s learning engine teaches 10 times faster, but the school’s promise to parents is not a full day of academics at that speed. The promise is twice as much learning in two hours, with the rest of the day returned for other forms of development. The product is called Timeback for that reason.

Puri presses the scale of the ambition and says he was “Alpha-pilled” less by the AI than by the 5K story: children and parents initially did not believe second graders could run a 5K, so Alpha trained them progressively — walking a quarter, walking half, jogging part of it — until they could. Liemandt treats that as a representative case. The point is not that every child becomes an elite runner; it is that children learn through experience that a hard goal can be decomposed and reached.

Joe Liemandt’s ambition is to reach a billion children in 20 years. He acknowledges that this cannot happen through Alpha-branded physical schools alone. There are about 100,000 schools in the United States, he says, and a billion children would require scale far beyond building campuses one by one.

Still, physical schools are the beachhead. Liemandt says the U.S. private school market is about $100 billion, enough in his view to create an anchor business with credibility and revenue even before public schools adopt the model. Shaan Puri compares this to Tesla’s Roadster: a high-end initial product that funds and validates the broader mission. Liemandt accepts the analogy.

Alpha, in this framework, is the high-end private school — “Stanford of K through 12.” He imagines a couple hundred locations, each with around a thousand K-12 students, but says that would still address only about 5% of the market. Around Austin, Alpha is incubating other school types to find product-market fit for different segments.

One is a gifted school, aimed at what Liemandt calls the 100,000 smartest kids, the subset who answer “more academics” when asked what would make them love school. Another is a sports academy, which he estimates may address 40% to 50% of the market because many children would choose sports if asked what they love. Liemandt says Texas Sports Academy’s baseball team won a national championship and beat IMG, that its basketball team is number one, and that its academics are “10 times better than IMG’s.” He also says Alpha is rolling out other sports with the goal of building teams that are both nationally elite and academically superior.

He also mentions Founders School, a high school model under development, and says different schools will use the same two-hour academic core while designing different afternoons. A Catholic school, for example, may want a different afternoon than Alpha does, but Liemandt believes many schools will still want the academic time compression.

He rejects the idea that AI education means children isolated with machines. In 20 years, he says, 90% of parents will still drop their children off at a building with other children and adults. Homeschooling exists, but he calls it a niche. The transformation he envisions is not the disappearance of the school building; it is a change in what happens during the six hours inside it.

To reach beyond physical campuses, Liemandt describes two additional paths. One is licensing or providing Timeback to people who want to build or transform schools, though he says he will not let schools use the software unless they are willing to rebuild around children loving school more than vacation. The second is outside-school products, including a “free to learn” video game being built by “some of the world’s best video game developers,” which he says will start shipping later in the year and is intended for 500 million children.

Joe Liemandt had largely avoided public attention for 20 years. He says 2005 was the year he got married and also the year of his last podcast appearance, at Stanford. He stayed quiet because being in the limelight was harder while building a family and because publicity no longer helped the business. Now, he says, publicity matters because education requires persuasion.

The biggest impediment to change in education, in Liemandt’s view, is parents. Not because they are hostile to their children’s interests, but because they are anchored to the system they experienced, as were their parents and grandparents. They do not believe the claims. They do not believe children can learn 10 times faster. They do not believe children would rather go to school than vacation.

Liemandt says Alpha’s marketing department’s slogan is “we educate the parents.” The phrase captures a strategic inversion: the school is not only selling to parents; it must first change what parents think is possible.

That is also why he believes the current moment is different from earlier education reform efforts. He says he spoke with a dozen billionaires before committing his own capital, each of whom, according to him, had spent more than a billion dollars on education. Liemandt says every one of them told him not to do it. They described education as the lowest-ROI philanthropic category, an impossible problem where even large sums do not move outcomes.

Liemandt’s answer is partly entrepreneurial self-belief — “I’ll be different” — but partly timing. Alpha existed before he committed fully. It had been started in Austin by MacKenzie and Brian, and Liemandt initially dismissed it as a “weird, janky school.” It took two years for MacKenzie to persuade him to visit. The turning point came when his daughters spent a week there before summer camp and came home saying they did not want to go to camp; they wanted to go back to Alpha.

For years, he thought the model was good for Austin families but not scalable. Generative AI changed that assessment. He connects the moment back to his high school paper on neural nets: “they’re finally here.” In his telling, AI made it plausible to scale Alpha’s academic engine and adult model to far more children.

He also argues that the capital model has to be different from philanthropy. He says he has put in a billion dollars from his software company, but says reaching a billion children will require tens of billions. Education is, in his description, a multi-trillion-dollar global industry and the second largest industry besides healthcare. He calls the opportunity “SpaceX for education.”

Nonprofits, in his view, do not scale well enough because they depend on donations. The better the product and the longer the waitlist, the more capital they need from donors to open another campus. He gives the example of high-performing charter schools with 3,000-person waitlists whose leaders cannot expand without a $40 million donation. Alpha’s model, he argues, can attract capital because demand from families can support new campuses.

The billion-dollar education commitment is funded by Joe Liemandt’s software-company cash flow. After Trilogy’s original configuration market matured and the dot-com downturn hit, Liemandt says 50% of Trilogy’s customers went out of business. The company discovered it was better at operating through the downturn than many other software companies, and distressed businesses were available “for a dollar.”

The acquisition strategy began opportunistically: buy a couple of cheap companies, see if Trilogy could turn them around, make money, then repeat. Liemandt says they have bought hundreds of companies since.

That strategy has now been reshaped by Alpha’s capital needs. Four years ago, Liemandt says he told his team they had spent 25 years telling him they were better than him, so now they could prove it. He would become, by his own description, their “worst shareholder,” pulling every dividend he could to fund education. The company had to shift from capital-intensive acquisition and turnaround work to a more capital-light model while still making acquisitions work.

He says the SaaS downturn and AI disruption have helped. Liemandt describes his company as one of the best operators of SaaS businesses and says it knows how to generate cash flow from them. In his account, many distressed SaaS companies are now owned by private credit holders such as Blackstone or BlackRock. Those creditors can either inject another $100 million or $200 million to reboot the company, or sell it to Liemandt’s company for a dollar and split the cash flows.

Four years ago, he says, creditors still hoped to rescue those companies themselves. In the last 12 months, AI has changed the calculation in his account; many no longer want to fund a risky AI reboot. Liemandt says the last pipeline he saw included $20 billion of companies being evaluated for cash-flow splitting and one-dollar acquisitions.

The acquisition machine, then, is not a side story. It is the financing apparatus behind the education project. Alpha is not presented as a conventional venture-backed ed-tech company. Liemandt says VCs will not fund ed tech because prior efforts failed, and billionaires have soured on education philanthropy. His answer is to use software cash flows to fund a private education platform until it can attract the scale of capital required.

Joe Liemandt is not arguing that children should simply use general-purpose chatbots. He says that if schools put chatbots on unmanaged Chromebooks in K-12, “those kids are not learning.” In that context, he calls chatbots “cheat bots” and says he understands why people want to ban AI.

His distinction is between unmanaged AI use and structured AI learning. Alpha’s model uses AI tutors and learning science to identify gaps, personalize instruction, and compress academic learning. The human adults are still central, but they are not primarily delivering academic lectures to a classroom.

He frames this around the future role of human expertise. Alpha uses a structure called a “Brain Lift,” which Liemandt also uses personally. The premise comes from a hierarchy he calls depth of knowledge: DOK 1 is facts, DOK 2 is summaries, DOK 3 is insights from facts and summaries, and DOK 4 is creating new knowledge. Large language models, he says, are good at helping with DOK 1, 2, and 3. Humans still matter most at DOK 4.

A Brain Lift is a curated body of reading, summaries, and insights that can be loaded into an AI system as context. Without that context, Liemandt says, asking an LLM to design a school will produce a teacher in front of a classroom. If he asks for a school without teachers doing academic teaching, he says, the LLM rejects it as unethical. But if he loads the Brain Lift — research papers, summaries, and prior insights — the model can reason from that context and help design Alpha-like structures.

For Liemandt, this is a core skill for the AI age: humans building expertise deeply enough to direct AI rather than accepting its default consensus. Alpha starts teaching this in high school. Personally, he says he spends about an hour a day reading experts, summarizing material, updating his Brain Lift, and trying to generate one DOK 3 insight a day. DOK 4 insights are rarer and larger.

He compares the practice to what he did in the late 1980s while building Trilogy: sitting in Stanford’s math and computer science library reading every paper on configurators, including work from Daimler in Germany that had to be requested and mailed. AI accelerates that process, but the discipline is the same: if he is going to enter a field, he wants to become expert enough to reason from the primary material.

His Warren Buffett study followed the same pattern. After the late-1990s boom, Liemandt says he once read Buffett and thought he was an idiot. After the 2001 bust, when his own returns looked much worse, he reconsidered. He read everything he could. His standard was that he should be able to pause a Buffett interview and answer the question the way Buffett would, because he understood how Buffett thought. When Buffett said something surprising, Liemandt would go back and investigate what he had missed.

That is the model of expertise he wants students to learn: not passive AI use, but deliberate knowledge accumulation, synthesis, and the creation of new claims that may initially sit outside what an LLM would produce by default.

Sam Parr pushes the implications of Joe Liemandt’s thesis with a deliberately provocative question: what happens when everyone has access to AI, everyone is smarter, and a billion people are improved by a significant amount? Liemandt answers by rejecting both naive optimism and dystopian flattening.

He says AI used wrongly can make children dumber. But used properly, he believes it expands the “human knowledge graph.” He gives the example of an Alpha student who, he says, is going to be published in Nature and will be the first high school student to be a lead researcher there. In his account, that kind of capability used to require being a Stanford PhD; now it can be pulled down to a 17-year-old. He says 10 Alpha students are behind her thinking, “she can do it, I can do it.”

At the same time, he does not claim education can erase all constraints. Parr asks how sports academies can make someone seven feet tall. Liemandt says they cannot. There are human limitations: height, IQ, working memory, age. He says his own neurons fire at about half the speed of his 18-year-old daughter’s, but he compensates with a knowledge graph built over more than 50 years. When she has to evaluate 10 things one by one, he can sometimes recognize that “seven’s the answer” because prior knowledge gives him shortcuts.

The point is not that everyone can become anything. It is that most people and parents misjudge the boundary. Alpha’s purpose, as he states it, is to get people to the edge of their actual potential rather than the much lower edge assumed by the current system.

That claim is why Liemandt presents education as the most important problem to work on. Raising the next generation is, in his view, the central job of every parent and every society. For a century, he says, there has not been a real answer for how to change schooling. Now he believes there is one — but it will require more than him. He says he needs to recruit “all the best people in the world” to the problem because it matters more than the alternatives.

By the end, he is also pitching education as a market. He tells the entrepreneurial audience that he has not yet fully proved the business model and go-to-market, but he intends to show that education is “the single best market” for entrepreneurs: a multi-trillion-dollar market with no competitors at the scale he believes is possible.