Summary

Liemandt’s core claim: Alpha School kids learn twice as much in two hours a day as standard schools deliver in six plus homework — verified on third-party MAPS tests — because AI tutors finally implement what learning science predicted for 40 years but couldn’t scale. The load-bearing number: “you can teach an entire grade level of material for one subject in about 20 to 30 hours,” versus 180 class-hours plus homework; his AP students all scored fives on AP World History spending “75 hours max.”
The product insight most edtech missed: “motivation is 90% of the solution. The edtech is 10%.” Edtech keeps failing because it serves the 5-10% who love academics for academics’ sake; Alpha’s #1 motivator is giving kids their time back — two focused hours buys four hours of workshops. And a categorical call: “90% of the use of ChatGPT is to cheat” — Alpha’s own AI has chat functionality deliberately disabled.
“All educational content is obsolete. Every textbook, every lesson plan, every test” — GenAI generates personalized lessons from each kid’s knowledge graph plus interest graph. The Timeback platform’s betas are hoped for between August and December of this year, with launch targeted next year, but Liemandt is candid about the gap: in tests with 50+ homeschoolers, “we’re not getting 2x in two hours” without the in-building motivational model.
The business framing for investors: education is “a trillion dollar market with no good products,” and “capitalism enters the chat” — he’s burning “hundreds of millions,” is putting in “the first billion” and 20 years, and argues nonprofit structure is why good schools stall (“I’m waiting for someone to give me $40 million… I have a 3,000 person wait list”). He’s also put $100M+ into AAA game designers building learning games, claiming “10 million moms in America who would pay a hundred bucks a month.”
Scaling runs through school choice, not districts: Texas’s new $1B school-choice program enables micro schools like Texas Sports Academy (a middle school per district for aspiring D1 athletes on the Timeback engine), GT schools, and entrepreneur-built models; public schools are “a decade from now.” The end state: within 5 years, “a tablet that’s less than $1,000 that is going to teach every kid on this planet everything they need to know in two hours a day.”
The funding engine is a dead-unicorn harvester: ESW/Trilogy buys failed SaaS companies from private credit “for a dollar” and splits the cash flows — he says $20 billion of those companies are out there on the books — via an installed-base playbook (call every customer, manage built-to-die DCFs). Forward view: pure transactional SaaS returns will be “dramatically worse,” while “AI native companies are going to be worth more than any SaaS company ever has been.”
The management OS underneath: Jim Abel’s three-lines-three-words strategy discipline (“create your strategy where there’s no escape”), Jack Welch’s brutal-feedback mentorship (“if General Electric doesn’t get a return on its investment, your product sucks”), and turn-it-to-11 conviction — people who copy “85% Joe” find “it doesn’t work.”

Deep dive

1. School rebuilt as a product — against a 200-year-old model

Liemandt’s frame is explicitly product-first: “I’m a product guy… how can you rebuild school and make it awesome?” The commitments are that kids must love school (the standard: prefer school to vacation — high schoolers recently asked to skip summer break), learn “10 times faster” via a two-hour AI-tutor session, and spend afternoons on life skills — leadership, entrepreneurship, financial literacy, public speaking — in team-based, non-screen workshops.
His history of the incumbent: the elite always had tutors (“Socrates, Plato, Aristotle, Alexander the Great”), but when universal education arrived ~200 years ago “we created the industrialized model” — a teacher in front of 20-30 kids, identical whether it’s a poor district in India or “a $75,000 private school” down the street.
Starting ~40 years ago, learning science (Stanford’s ed school among the leaders) showed the lecture classroom is “the worst way to teach kids.” Bloom’s two sigma is the seminal paper: individualized tutoring produces dramatically better results but was “obviously not feasible.” Mastery, spaced repetition, cognitive load theory — all known, none scalable. His own arc: an 80s expert-systems guy (“I wrote a paper in high school on AI… neural nets, but they’re decades away”) who saw GenAI three years ago and concluded “this is finally the technology that is going to allow education to scale” — and became principal because “I got to be in it.”
Even he needed convincing: “Every parent wants to educate the kid the way they were.” Mackenzie spent two years recruiting him — “no offense, Mackenzie, I’m not sending my kid to your weird janky school.”

2. AI is learning science’s microscope — the closed loop nobody else has

His favorite analogy: biology, chemistry and physics wandered “lost in the wilderness” — bloodletting doctors, self-detonating chemists — until an instrument (microscope, analytical balance, telescope) enabled precise measurement. Learning science has been in its wilderness for 40 years because classroom effects drown the signal: “the environmental variables dominate.” AI enables precise instruction plus closed-loop measurement — generate a lesson, measure time spent, test on standardized instruments, iterate. “We’re basically the only place that has that closed loop, which is why we’re able to get these crazy awesome outcomes and other people aren’t.”
The mechanics start with separating age grade from knowledge grade: an assessment maps what you know, then lessons build from there — a sixth grader struggling with fractions gets the fourth-grade division he actually missed, because “most of knowledge is hierarchical… algebra is basically advanced fraction manipulation.”
Mastery means >90%, not the passing-at-70% standard — his basketball analogy: a point guard losing the ball 30% of the time doesn’t get alley-oop practice, “let’s get back to the basics.” He warns that off-the-shelf apps embed a mastery standard that’s “just too low” to replicate Alpha’s results.
The cram experiment kids run on themselves: one group does vocabulary 10 minutes a day, another crams the night before; both score the same on Friday, then a retest a week later shows “the cram kids don’t remember anything.” Cognitive load theory in practice: working-memory “slots” (registers on a CPU, for the technical), reps to move knowledge to long-term memory — different per kid.

3. The numbers: a grade level is 20-30 hours

The stat he leads with for parents: a kid a year behind in a subject is only 20-30 hours behind — “there’s 180 school days and each class is an hour… and you’re telling me it’s 20 to 30 hours. And you’re like, yeah, 5 to 10 times faster.” Two years behind is 40-60 hours, which can be caught up with an hour of homework tracked on a parent dashboard.
Proof points he cites: AP World History students “all got fives” spending “75 hours max”; January MAPS midyears (third-party) showing kids “learning twice as much in two hours a day” — timed to rescue parents from the skeptical-relatives-at-Christmas wobble. Growth works both directions: “you can take a kid at 50th percentile and get them into top 10%. Or you can take a kid at 99th percentile, which is usually capped out, and all of a sudden put them years ahead.”
The signature fluency story, offered with a hedge (“an extreme example and I hesitate to say it”): a student with a 740 math SAT sent back to third-grade math because missing multiplication-table fluency was overloading working memory into careless errors. “We’re the only school in the world who will take a 740 math SAT student, go back to third grade — and she got a 790.” Parent and student hated the prescription; “but it is” the issue.
His response to that expert view: experts say “theoretically this has all been predicted for 40 years but no one’s been able to implement it. Oh, okay. I guess we’re implementing it and it’s working.”

4. “All educational content is obsolete” — and the engineering behind generation

The thesis quote: “All educational content is obsolete. Every textbook, every lesson plan, every test” — because GenAI generates lessons from three inputs: the curriculum, the student’s knowledge graph, and their interest graph. The first-grader who “hates books” gets a choose-your-own-adventure at exactly his Lexile level starring him and his soccer buddies saving the world; the reluctant math student who wants to be a fashion designer learns fractions “in rulers and dress patterns.”
The examples as told: students back from a Taylor Swift concert struggling with the start of World War I got a generated table mapping the award-show controversy onto likely Archduke Franz Ferdinand — “the kids are like, okay, I’m going to remember this forever.” High schoolers used Sumo (AI music, “just like Hamilton”) to generate a song per AP History unit, since a transfer student’s history knowledge “is basically whatever is in a Hamilton song.”
Lessons hold 80-85% accuracy — the zone of proximal development — “talk to game developers… you use the same theory as game developers.” The engine, Incept, is a standard AI team “embedded with massive a set of learning scientists” — “probably the biggest, best team of learning scientists” — which he argues is why generic LLM education products fail: no mastery standard, so kids just cheat. The name is deliberate: the movie reference is a mind-virus, plus the original Oxford/Cambridge meaning, “qualified to teach” — trust matters because “you have TikTok as an untrusted generator that currently dominates… we have to out-TikTok TikTok.”
The second AI: a vision model streaming each kid’s screen, hunting “anti-patterns” — skipping explainer videos, scrolling to search answers — bootstrapped three years ago by humans annotating videos all night, “like the data stream of all the Tesla cars.” Its waste meter is the expensive part (“you’re streaming full-time”), and kids embrace it: “you’re wasting 50% of your time — and the kids are like, whoa, coach me up.”

5. Motivation is 90% of the solution — and time back is the motivator

The central claim: “motivation is 90% of the solution. The edtech is 10%.” Edtech has failed (“we keep putting edtech into schools and test scores keep going down”) because it’s necessary but not sufficient — a paper he cites, “the 5% solution,” shows edtech wins concentrate in self-motivated, high-conscientiousness kids who’d thrive anyway. “That’s not 90% of the market.”
The founding negotiation, as told: new fifth graders said they’d love school if there were “no school.” The deal struck — two hours of genuine engagement buys four hours of awesome — became the product: “the code name of our product… was Time Back. Give kids their time back. And that is by far the number one motivator for kids.” The other motivators work once time back is solved.
His most categorical tech take: “You are not a principal at a school if you think ChatGPT is good. 90% of the use of ChatGPT is to cheat” — because kids want their time back. Alpha’s AI literally has chat disabled; the right uses are lesson generation and screen-watching coaching, not chatbots.
Gamification is Apple Watch-style rings closed by 25-minute pomodoros at 80%+ accuracy; new transfers start in “standard school” — literal desks — and graduate to bean bags and privacy booths by proving they’re two-hour learners. Motivators are individualized down to the campus: at Brownsville, “the biggest motivator for the youngest kids is the petting zoo… they’d go do 10 hours of homework” for the bunny.

6. The afternoons — and the claim that high standards are the key to happiness

His self-declared “most controversial thing”: three beliefs he wants every American parent to hold — kids must love school; kids can crush academics in two hours (“you cannot subject your kid to six hours a day plus homework”); and “the key to your child’s happiness is high standards.” Kids “want to struggle and fail on their road to success,” supported by caring adults — high support, high standards.
Life skills are taught deliberately rather than outsourced: “today teamwork and leadership is taught via sports in America.” Fifth graders ran an Airbnb (learning P&Ls and pricing), then launched a breakfast food truck — “Dippers” — because, as one told him, “do you understand the gross margin on breakfast foods? It’s just enormous”; they clear $1,000 on weekends at car dealerships. Public speaking is framed as the postgame press conference, ending with kids being interrogated by real press at Austin FC.
The 5K story: second-grade guide Faith committed every second grader to running a 5K under 35 minutes — Liemandt himself flinched (“that seems too hard, and God, the parents are really not going to like that”). She taught them Atomic Habits (“most adults don’t learn that till after college”), walked the full track on day one to establish everyone finishes, then incremented to the December Jingle 5K. The payoff line kids brought home: “Mom, dad, I know how to do anything now.”
High school flips to self-direction: the four-year “AlphaX” passion project — one student is building the first all-teen-produced Broadway musical, casting singers off TikTok, able to fill a thousand-seat venue. “There’s no guide or coach at Alpha who can help her” — the discovered pattern is that “super successful adults overinvest in ambitious teens,” so kids learn to DM for help.

7. Guides, not teachers — $100k minimum and zero academic teaching

No adult at Alpha does academic teaching — hence “guides and coaches,” not teachers, though he concedes the terminology is unsettled. The accountability standard is audacious: every eight weeks kids are surveyed on whether their guide is the one-or-two-teachers-who-transformed-your-life figure. “The reason they transform your life is not because they graded your seventh grade science quiz really well.”
Minimum guide pay is $100,000 — “basically double” the Austin market — which opens the talent pool beyond credentialed teachers: ex-NBA coaches, ex-NFL and MLB players at the sports academy, hired for motivation and high standards. He cites the book “10 to 25” by Dr. Joerger as the operating philosophy.
Guides work a motivation database — “every motivation that’s ever been tried by any guide at Alpha” per kid — and pull students out of pomodoros for one-on-ones. Ratios run ~5 students per guide when young, thinning by middle school because adolescents “want more freedom, but the ones they want, they want really high caliber.”
Receiving feedback is itself a tested life skill, from “can you receive feedback without crying” in kindergarten to actively seeking critical feedback by eighth grade — modeled on athletes “watching game film,” applied to test errors: careless mistake or didn’t-know-it?

8. Money works — $100-for-100 and breaking the IQ link

The systemic indictment first: American learning curves flatten in middle school — “the median person in high school in America actually doesn’t learn anything… on a MAPS test, which is a 300-point scale, the median American goes up one point in four years.” Cause: time-based rather than mastery-based education — Saul Khan’s Swiss-cheese foundation that eventually “collapses and you can’t learn anymore,” usually at algebra.
His hole-filling hack, told in full: seventh graders offered “$100 for 100” on the Texas Star — any grade level. They start at third grade, collect $100, ladder up until they hit a 75 or 85, accept AI-generated lessons to close the gap, and keep going. “For literally $400, we can hole fill… but second, you actually change how they think.” Fewer than 10% of Alpha parents believe a kid can score 100; after a year, 90+% of kids believe it — and his K-6 building “gets more hundreds on the Texas Star than a school district of a hundred thousand kids.” The reframe: “It’s not about IQ. It’s just about effort. And as soon as you break that link for a kid, it changes everything.”
Alpha Bucks route earnings into things parents endorse: funding passion projects, and investment accounts for middle schoolers — real earned money (“because your loss aversion”), with a 50% early-withdrawal penalty parents love. Context on spend: the U.S. spends ~$20k per kid (Alpha $40k+), “we spend plenty of money” — a trillion dollars on K-12.
The story he tells on his own daughter: convinced she wasn’t her academic older sister, she took his $1,000 bet to reach top 1%, filled an Amazon cart, stared at it each morning, and ground the apps — over his wife’s objections. Her verdict to her mom afterward: “I never thought I was as smart as my older sister. I now realize I can do anything.”

9. Scaling: Timeback, micro schools, and $100M of video games

Honest status report: “the product’s not done” — hallucinations persist (“the kids take a screenshot and send it in, like, AI’s acted up today”), and he hopes to have betas between August and December of this year, with the platform targeted for next year. Bigger admission: with 50+ homeschool testers, “we’re not getting 2x in two hours” — because the building supplies the 90% motivation layer that a kitchen table doesn’t. He believes a working homeschool version will be available next year.
Distribution runs through school choice: Texas just passed a $1B school-choice program, enabling Texas Sports Academy — a planned micro middle school in each of Texas’s ~1,200 districts for aspiring D1 athletes, academics on Timeback in the morning, a revered local coach all afternoon, launching when ESA vouchers come due a year out. Plus GT schools for the 5-10% who “love academics for academic sake” (his oldest read calculus books for fun; those kids learn “3x, 4x, 5x” and joke they’ll be “the last human smarter than AI”). Public schools are “the very last in the adoption curve” — his vision needs all 20 years.
The video game bet is his scalable-motivation play: “the people who know how to engage kids more than anything is video game developers.” He’s invested over $100 million in AAA designers in Austin building a game on the two-hour engine “that they believe will rival any video game on the planet” — plus a Candy Crush-like for math-fact fluency. The tell from testing: asked how long they’d grind a “training center” of math problems for extra in-game health, one kid said 10 hours — “remember, it’s only 20 to 30 hours to do an entire grade level.”
The endgame he pitched to foreign education ministers: within 5 years, “a tablet that’s less than $1,000 that is going to teach every kid on this planet everything they need to know in two hours a day. That’s coming.” Alpha’s AI burn is ~$10,000 per kid now, “but it’s going to be down.” His asymmetry warning: AI is bad for Uber drivers, but for kids “it is an unequivocal good” — “this is the best time in history to be a 5-year-old.”

10. The business of education — and the hardest skeptic questions

“Capitalism enters the chat”: he argues education must support real businesses to attract talent — “it’s a trillion dollar market with no good products.” The nonprofit structure is the scaling trap: a good school system or charter school sits on a 3,000-person waitlist “waiting for someone to give me $40 million so I can build a campus,” whereas a business model means “capitalism will provide capital.” He’s in the burn phase — “hundreds of millions” — and Mackenzie’s expansion needs $300M.
The skeptic questions he concedes are hardest: “we haven’t done this at scale,” plus infinite money and selection effects. Deeper still: parents don’t agree on what education is for — in Austin, parents defend academically failing schools because one teacher “transformed their life… I don’t want academics.” His confession: “one of the things that I did not realize going in is how few parents actually care about academics.”
The positioning lesson took “hundreds of parent meetings”: pitching “2x learning” got pushback (“why are you pressuring him to learn 2x?”); the identical product pitched as “two-hour learning” flipped the same parents into allies (“Johnny, don’t you want to do the cool things?”). “Positioning matters” — and different families want different bundles: the sports-academy mom hoses her kid off after practice while the GT family continues the chess tournament at home.
His forward call on software investing: pure transactional SaaS returns will be “dramatically worse” over the next 20 years, but “AI native companies are going to be worth more than any SaaS company ever has been” — because they change how the business works, education being the proof case of a multi-trillion-dollar global industry where nothing previously built could make kids learn 10x faster.

11. Joe’s own story: Trilogy, dead unicorns, and learning at the limit

The backstory, mostly absent from the public record since the ’90s: hated school (“what’s the minimum academic score I can get” — he once submitted a test unfinished at exactly 89.5 because it rounded to an A), dropped out of Stanford after studying under expert-systems father Ed Feigenbaum, and built Trilogy — “the first AI product to sell a billion dollars,” a configurator for phone switches, Boeing airplanes, and medical equipment — where selling it wrong was only discovered on the manufacturing floor and cost millions to fix. Trilogy University recruited 2,000 Ivy grads to Austin with “the hardest 100 days you’ve ever done” — beating peak-Microsoft 70/30 for graduates because it was harder. The ethos survives as Gauntlet AI: 100 days of 100-hour weeks, 50%+ washout, $200k jobs waiting.
Today’s cash engine is the dead-unicorn trade Bill Gurley called for: ESW/Trilogy takes over-levered, zero-equity SaaS companies from private credit — some “basically for a dollar” — and splits the cash flows, and says $20 billion of such companies are out there on the books. The playbook: flip from growth to installed base (“the first thing they do is call all the customers, every single one”), accept built-to-die economics, manage the DCF as the ice cube melts. That cash is why “you can take a billion dollars and say I’m going to go spend it on education.”
The best business advice he says he ever got came from HR chief Jim Abel: strategy must be three lines, three words each. Liemandt ignored him (“smart person disease”); Abel interviewed every employee and whiteboarded what they actually thought the strategy was — “literally the opposite.” The principle: “create your strategy where there’s no escape,” and marketing must be “a magnet — attract and repel”; if nobody would say the opposite (“integrity”), it’s a platitude. “Kids must love school” is built to pass the Abel test.
Jack Welch, his late father’s boss, mentored him after his dad died when Joe was 25 — brutal feedback included: “I don’t care if it comes from heaven above. If General Electric doesn’t get a return on its investment, your product sucks.” And on his own reputed “one-tenth business killer psycho”: “pushing the limit… is where you get all the alpha.” People ask for “85% Joe” and find it fails — “I’m pushing it to where it needs to be for it to work… people are like, that standard is so ridiculous. And you’re like, yes, that’s the one that makes it work.” (Elon Musk at dinner, for calibration: “I don’t understand why you don’t have a thousand schools already.”)

Google Part II: Alphabet (Audio)Microsoft Leaders on AI-Driven Commercial and Advertising Growth