Peter Attia MD366 ‒ Transforming education with AI and an individualized, mastery-based education model
CHAPTERS
- 0:00 – 6:33
Joe's path from Stanford dropout to Trilogy
Peter Attia explains why an education conversation fits a health podcast, then asks Joe Liemandt about his background. Joe describes growing up on the East Coast, writing a high school paper on AI, and leaving Stanford before senior year to start Trilogy. He explains the time-to-market pressure behind that decision and the move to Austin.
- •Education matters for health because today's students will be tomorrow's caregivers
- •Joe's father worked for General Electric, so the family moved often
- •A high school AI paper and a Stanford class on expert systems led Joe to drop out
- •He felt a time-to-market race, though the product took over three years to build
- •The founders lived in a garage, then moved to Austin because it was cheaper
- 6:33 – 12:38
Discovering Alpha School and its early apps
Joe recounts how McKenzie Price's Alpha began in a garage, how he resisted it, and how his daughters' reaction won him over. He describes the early school's rules, its small mixed-age setup, and the weak educational apps of the time. His daughter's failed seventh grade standardized test exposed gaps in those apps.
- •Joe's daughters asked to keep going to Alpha instead of summer camp
- •Co-founder Brian's two beliefs: kids must love school, and kids can do far more than expected
- •Early Alpha had about 20 kids, apps for academics, and life skills workshops
- •Older apps had low mastery standards and limited use of learning science
- •Fixing the gaps in those apps is what pulled Joe into education
- 12:38 – 20:04
Why U.S. academic performance keeps falling
Peter raises the roughly $1 trillion spent on K-12 against poor results, and Joe answers with data. He cites falling norms on a national test, easier AP standards, and placement tests at Alpha showing that A and B students are years behind. He argues that missing fundamentals drive the problem.
- •Average eighth-grade knowledge was lower in 2025 than 2020, and lower in 2020 than 2015
- •Only the 99th percentile group has improved
- •Alpha's placement tests found A students up to three years behind and B students three to seven years behind
- •Joe says dropping multiplication-table memorization overloads working memory
- •A 740 math SAT student rose to 790 after going back to third-grade multiplication and division tables
- 20:04 – 28:38
Mastery versus time-based progression and the 10x claim
Joe argues that a time-based system moves kids on without mastery, which leaves gaps in prerequisites. He uses sports analogies and the hierarchy of math and science knowledge to explain why. He says AI tutors and apps let students finish a grade level in 20 to 30 hours, and cites Bloom's two sigma research.
- •Chemistry and physics build on algebra and calculus, and algebra builds on fractions
- •Peter recalls a student failing chemistry because of unmastered fractions
- •Teachers face students with an age grade that differs from their knowledge grade
- •Alpha kids finish a grade level of a subject in about 20 to 30 hours
- •Joe says about 95% of eighth graders could reach what is now top-10% performance in math
- 28:38 – 38:59
Timeback and motivation as the key to learning
Peter asks why scores fall even though schools were never mastery-based, and Joe points to lowered standards and parents who don't prioritize academics. Joe explains how Alpha reframed its pitch around two-hour learning, creating the Timeback model. He says motivation is 90% of the solution, and that the AI coaches students and keeps them in the right difficulty zone.
- •Joe says fewer than 10% of parents really care about academics, so Alpha pitches two-hour learning
- •Timeback means students earn the rest of the day by engaging with the apps for two hours
- •The AI watches the screen and coaches students instead of acting as a chatbot
- •Zone of proximal development: keep students at roughly 80-85% correct
- •Joe says lecture retention is about 5%, and current engines teach 20-40 facts per hour versus 1-5 in a Stanford lecture
- 38:59 – 46:09
Learning science consensus and high standards
Peter asks what most learning scientists would agree on, and Joe lists individualized mastery tutoring, spaced repetition, the testing effect, and cognitive load theory. They connect fluency to elite sports performance. Joe then gives the three beliefs he asks Alpha parents to hold and explains the struggle, fail, support cycle.
- •Learning scientists broadly agree on individualized tutoring with mastery standards
- •Peter compares cognitive bandwidth in Tom Brady and Formula 1 drivers to math fluency
- •Joe's three parent beliefs: kids must love school, they can learn in two hours, and high standards bring happiness
- •Parents' wish to protect kids from struggle drives low standards
- •Joe says mastery removes IQ and conscientiousness as limits on K-8 math
- 46:09 – 52:56
Breaking the "I can't do math" story
Peter notes that many students build a narrative that they can't do math. Joe says no learning science supports the idea that middle school girls can't do math and science. He gives examples of Alpha students with very high SAT math scores and says that in a mastery system, achievement becomes a decision rather than an inherent ability.
- •Joe says there are no papers showing middle school girls can't do math or science
- •Seniors with arts and TikTok interests scored 790 on SAT math
- •Filling prerequisite holes unlocks algebra, then chemistry and physics
- •Freshmen average 1410 on the SAT by year end, according to Joe
- •His younger daughter asked the AI how long it would take to reach a perfect 800
- 52:56 – 1:00:38
Using money as kindling to unlock student belief
Joe explains that Alpha uses extrinsic rewards such as paying middle schoolers $1,000 to reach the top 1%. Peter asks why people feel uneasy about this, and Joe describes the intrinsic-versus-extrinsic debate. Joe tells the story of his daughter's Amazon cart. Peter calls the reward kindling that starts a self-sustaining fire.
- •Alpha pays middle schoolers $1,000 to get to top 1% academically
- •Joe says the larger payoff is a change in how kids see themselves
- •Some papers say payment undermines intrinsic motivation, but Joe says Alpha's kids weren't motivated to begin with
- •His daughter filled an Amazon cart as motivation and later said it changed her view of herself
- •Peter describes the reward as kindling that doesn't have to continue forever
- 1:00:38 – 1:05:52
$100 for 100: filling gaps with a simple incentive
Joe says the simplest fix for American education would be paying kids to fill their gaps. He tells how he offered seventh graders $100 for a perfect score on a state test from any grade level, starting at third. Peter notes how this plays out with the 90-versus-100 mastery threshold.
- •No student wants to go back to fourth grade, and parents resist it too
- •Students earned $100 bills for 100% on third- through seventh-grade state tests
- •The AI generates lessons for the questions each student missed
- •Alpha's mastery standard is above 90, and the 100-for-100 offer pushes kids toward perfection
- •Fewer than 10% of incoming parents say their kid could score 100, versus 90% of kids after a year at Alpha
- 1:05:52 – 1:19:49
AI as education's light microscope
Peter and Joe compare AI's role in education to the light microscope in medicine. Joe says AI offers precise teaching and closed-loop measurement, so learning science can finally be tested and applied. He explains that classrooms will still have adults and other students, and describes how generative AI can personalize lessons around each kid's interests.
- •Learning science has existed for 40 years, but there was no technology to implement it
- •AI gives exact control over what is taught and measures what each student knows
- •Timeback's math curriculum is expected to teach in 20% less time this year
- •Teach Tales builds choose-your-own-adventure reading around a child's interests
- •Joe says current content is static with human review because of hallucinations, with dynamic generation expected in 2026
- 1:19:49 – 1:29:32
Joe's $1 billion bet and the path to mainstream adoption
Joe explains that generative AI's arrival three years ago prompted him to step back at Trilogy and put a billion dollars into education. Peter asks when Alpha's model must merge with the existing system. Joe describes the obstacles: parents who value community over academics, skepticism toward ed tech, and teachers who aren't trained in learning science.
- •Joe took a billion dollars out of Trilogy three years ago and plans a 20-year commitment
- •Joe says a billion dollars isn't enough for a trillion-dollar industry, and expansion needs hundreds of millions in capital
- •Skepticism toward tech billionaires fixing education is well earned, in Joe's view
- •An Austin example shows parents defending failing schools for community reasons
- •Joe says teachers aren't taught learning science, and about 80% would welcome the guide role
- 1:29:32 – 1:33:56
Guides, the rebuilt school day, and costs
Joe describes how guides, who are often former teachers, spend their time on motivation and mentorship instead of lecturing. They discuss the two-hour learning block plus four hours of life skills and workshops, and why superintendents struggle with that redesign. Joe covers current AI costs of about $10,000 per kid and the goal of a sub-$1,000 on-device tablet.
- •Guides use 25-minute Pomodoro blocks and one-on-one check-ins
- •Afternoons are for workshops on leadership, public speaking, entrepreneurship and financial literacy
- •Superintendents ask what kids do the rest of the day, since the model rebuilds the whole school day
- •Current AI costs are about $10,000 per kid per year
- •Joe expects on-device AI and a sub-$1,000 tablet for the K-8 core within three to five years
- 1:33:56 – 1:44:08
Reimagining school across five dimensions
Joe lays out five dimensions for rebuilding school: kids love school, learn 10x faster, develop life skills, have transformative guides, and build character, community, classmates and culture. He gives examples such as second graders running a 5K and the high standards, high support parenting framework. He then covers the lower-priced schools and scholarships.
- •Kids must love school more than vacation, and about 50% say they do
- •Life Core is Alpha's common-core-style curriculum for life skills
- •Second graders worked up to a 5K to learn the 1% better habit
- •The mentor mindset framework from Dr. Yeager calls for high standards plus high support
- •Alpha costs $40,000 to $75,000, with scholarships and cheaper school models
- 1:44:08 – 1:48:58
Closing the wealth gap in academic outcomes
Peter reads statistics on achievement gaps between rich and poor districts and asks whether they could be erased. Joe says the gaps could be compressed. He explains that wealthy families buy one-on-one tutoring and that AI can provide individualized learning cheaply. He suggests credible institutions could build an open-source solution.
- •Peter cites gaps such as four grade levels by sixth grade and a 77x Ivy League disparity
- •Joe says the gap could be totally compressed
- •Wealthy parents pay for after-school tutors to fill holes the school leaves
- •Post-COVID high-dosage tutoring worked when one-on-one and faded when moved to six-to-one
- •Joe floats an XPRIZE-style open-source effort led by respected institutions like Stanford
- 1:48:58 – 1:54:30
Cultural adoption, parent buy-in, and school variety
Peter asks about the biggest risk, and Joe says it is cultural and systemic rather than technological. The tutor software will not work if kids sit in class for six hours, and parents need a trusted reference and a visible result. He describes Alpha's varied school types and the growth plans.
- •Joe says the real challenge is rebuilding the school system, not dropping in ed tech
- •The software wouldn't work if kids still sit in class six hours a day
- •Parents need a trusted reference and to see their kid do something new
- •Alpha has gifted and talented, sports academy and wilderness schools, with arts and music planned
- •Joe expects 25 Alpha campuses and 25 non-Alpha campuses, and hundreds more within 12 months
- 1:54:30 – 2:03:22
Peter's daughter's experience and a call for talent
Peter shares his daughter's two years at Alpha, where she overcame a negative story about math and built a strong foundation. Joe thanks the family for their founding-family feedback. They close with Alpha's expansion and Joe's appeal for engineers, entrepreneurs, educators, learning scientists and capital.
- •Peter's daughter entered seventh grade convinced she couldn't do math and was acing it by eighth grade
- •She later returned to public high school and has continued to do well
- •Her bond with her guides was strong enough that she cried at eighth grade graduation
- •Joe says the family's feedback helped improve the lessons
- •Joe invites people to email him with 'Heard you on The Drive Podcast'
