a16z Just Launched a School With OpenAI, NVIDIA & Anthropic

Summary

• Gagan Biyani is launching Andreessen Horowitz Academy — an unaccredited, residential, post-high-school program in the heart of San Francisco, backed by $40M raised deliberately for “a few years” of runway rather than 18 months. His reasoning: “there’s a big difference between starting a school and starting an AI app” — students need to know the institution won’t run out of money, and he thinks the raise is enough “to get to profitability on the school.” The promise is to develop AI-native builders with super-high agency and the people skills to understand Silicon Valley.
• The business model is unapologetically for-profit with three revenue streams: tuition, corporate partnerships, and equity connected to businesses students join or start. The founding class fellowship (September 2027, one year, 50 students, applications open now) is tuition-free apart from room and board; the two-year program planned for 2028 pending regulatory approval will charge “equal to or more than the top-tier private elite universities” on the bet that it’s “a high-value institution, not an elite institution.”
• Biyani says founding partners include Anthropic, Coinbase, Meta, and Replit, and calls NVIDIA a core partner; Molly also names Anduril, OpenAI, Palantir, Stripe and others. The wider partner network supplies compute credits, curriculum input, and hiring connections. Biyani’s structural argument for why this beats universities: “the leading researchers in AI are no longer at university campuses, they are at these companies,” so the academy can offer “front-row seats” to leading AI companies.
• His most contrarian call is against the young-founder funding frenzy: VCs are giving young builders “$150,000 to $250,000 checks on a flyer hoping this is the next Michael Truell,” but “only a very, very, very small percentage of these builders should be starting companies right away.” That’s the wedge versus the Thiel Fellowship, YC, and a16z SPEEDRUN, which are generally oriented toward people starting a company now — “startups are a 10-year slog and most of them don’t work out… you shouldn’t just take money because it’s there.”
• Pedagogy is built around the “rise of the super IC”: students learn to “orchestrate 30 agents at a time… agents that are orchestrating agents,” with 75–80% of time on self-chosen projects and classes that are “didactic only.” Admissions run on proof of work, not grades — the internal filter phrase is “you did what?”, exemplified by a 16-year-old in Tennessee building a biology LLM to design molecules, later pivoting to proteins.
• On reports of students booing AI at commencements and a friend’s report from Columbia, Biyani is unbothered: “we want the AI boomers,” and self-selection is the feature — most students should not fit a 50-person class. He says AI is here and coming more strongly, and the academy will embrace it by giving students resources to build with AI, subject to limits. The eventual scale could be 500 or 1,000 students, but he does not know yet.
• The talent-pipeline thesis is concrete and trackable: Alpha School students are now starting at Berkeley and Stanford, and Biyani expects the transition from a high-agency environment to college to be difficult — “I am going to be tracking this like a hawk.” The addressable market is the builder-misfit archetype nationwide, screened through portfolios, interviews, and planned IQ, personality, and writing tests, judged on “slope and not experience.”

Deep dive

1. A school built like a startup: unaccredited, experiential, and residential in San Francisco

• The announcement: Andreessen Horowitz Academy — Molly’s reference to a16z being in “reverse order” — is a post-high-school residential program for young builders who want jobs or companies in tech. Biyani calls it “a completely new take on what you could do post-high school in the AI era” and a chance to show “a positive and optimistic frame on what college could look like if it adopted the principles of an AI-native education.”
• The structural break from college: completely unaccredited (paid programs still need regulatory approval), and project-based rather than credit-based, “because in the AI era, you can learn whatever you need to learn on the job.” Three project categories — research, learning, and building — are illustrated by a hypothetical student visiting factories in Shenzhen or Austin while studying robotics online and building “a new type of robotic vacuum” or playing with “the new Hugging Face duck.”
• Biyani’s stated promise is that students will become AI-native builders with super-high agency and the people skills to understand Silicon Valley. If the academy can deliver all three, he believes it could be the most valuable educational experience anyone could have.
• Location is non-negotiable: not Alcatraz, and perhaps not Treasure Island despite the memes — “this has to be in San Francisco… in the heart of San Francisco. We want the students to be exposed to the city itself.”

2. $40M raised for longevity — and a spiky bet that the school reaches profitability

• The raise was sized against institutional survival, not startup convention: “instead of raising for 18 months, we raised for a few years… as a student, you have to know that this school isn’t going to run out of money.” Investors include Adam D’Angelo, Tobi Lütke, Tony Xu, Fidji Simo, and Palantir’s Shyam Sankar.
• Molly’s direct question — how do you get to profitability on a school? — draws the model: “this is a big, spiky point of view, but we believe this school is going to be so valuable that we’re going to charge for it.” Tuition is expected to be the bulk of revenue; corporate partnerships could become significant because companies want access to strong students; and over time, Biyani hopes to monetize through equity connected to businesses students join or start.
• The founding class fellowship starting September 2027 is one year and tuition-free, though students cover room and board. The two-year program planned for 2028, pending regulatory approval, will carry tuition “equal to or more than the top-tier private elite universities” — Biyani’s framing being “a high-value institution, not an elite institution.”

3. The partner roster and company access

• Biyani says his initial outreach to Anthropic, Coinbase, Meta, and Replit led to founding partners; he calls NVIDIA a core partner. Molly names a broader partner list including Anduril, OpenAI, Palantir, and Stripe. Separately, Biyani says there are 30-plus hiring partners, including Shopify.
• Their contributions — still being finalized per partner — span curricular advice, feedback on the talent companies want, and compute credits, which Biyani flags as an important constraint: young people can build things no prior generation could, but “that access comes at a cost.”
• The load-bearing argument: “the leading researchers in AI are no longer at university campuses, they are at these companies” — so the relationships could provide “front-row seats to whatever is happening at the leading AI companies.” For NVIDIA specifically, Biyani plans to ensure students get access to its latest knowledge about the future of AI.

4. Admissions runs on “you did what?” — proof of work over grades

• The peer-quality promise is central: “we are not going to prioritize your grades. We’re going to prioritize proof of work.” The internal catchphrase from student research is “you did what?” — exemplified by a Tennessee student who, at 16, built “an LLM for biology that is beating some of the leading models,” initially designing molecules before pivoting to proteins because he decided that was the market to pursue.
• Selection targets motivation, not credentials. Biyani’s Illinois Mathematics and Science Academy anecdote: a student told him “almost all my fellow students are doing extracurriculars solely for the purpose of getting into college. And I do extracurriculars because I enjoy it” — a difference “not obvious on a piece of paper.” The academy plans to interview students and dig into what they have built.
• Mechanics: portfolio submission first, followed by an in-person interview whose second half is planned to include IQ, personality, and writing tests, judged on “slope and not experience.” A USAMO participant who only discovered building in senior year could still fit. Even applicants who do not get in may benefit: the process “will be its own education in what they could do to optimize their college experience.”

5. The retention problem — and why most teen builders shouldn’t take VC money yet

• Molly’s sharpest pushback: with students this good, VCs “are going to be flocking around trying to grab these founders… and just fund them with money” — so how do you keep them? Biyani concedes it’s real, framing today’s builders as caught between “two very awkward options”: a four-year university “not designed for building,” or investors giving out “$150,000 to $250,000 checks on a flyer hoping that this is the next Michael Truell” of Cursor.
• His answer is a release valve, not a cage: “if you have an incredible idea… we will support you” and help find investors; students can leave “in two months, three months.” But for the vast majority, one to two years of exploration beats chasing “the first decent idea you have” — “startups are a 10-year slog and most of them don’t work out… you shouldn’t just take money just because it’s there.”
• Versus the Thiel Fellowship, YC, and a16z SPEEDRUN, which are generally about starting a company, the academy defends “the exploratory nature of a post-high-school academic experience” — cohort camaraderie, colocated housing, and mentors spanning crypto, robotics, AI applications, classic SaaS, and other fields.

6. Doomers need not apply: the super-IC thesis and the calculator test

• Molly cites students booing AI at commencement ceremonies and a friend’s report from Columbia, where AI was booed when mentioned during the school year. Biyani’s response: “a school that embraces AI is a school that embraces reality and the future… we want the AI boomers.” Self-selection is fine because only 50 students need to fit initially — “maybe eventually we’ll get to 500. Maybe we’ll be at 1,000. I don’t know.”
• The pedagogical core imports Maven’s thesis: middle-management orchestration of humans is giving way to “the rise of the super IC,” so students should “learn how to orchestrate 30 agents at a time. 50 agents… agents that are orchestrating agents” — with the goal of being “one of the highest performers at these companies,” not merely employed.
• His nuance on AI dependence, via the calculator analogy: he has one, but if he cannot read an employee’s spreadsheet and see how a 5% conversion improvement hits the P&L, “I have no business being CEO of this company.” At Maven he requires a human-written first draft; the academy will not impose that rule — “the school is going to be much more open… results-oriented” — but cohabitation with AI requires understanding what it is doing.

7. Curriculum design: 75–80% building time, didactic classes, and weekly pod accountability

• The inversion of college: preserve “75 to 80% of students’ time for doing work on projects,” with classes on product sense, AI architecture, negotiating, and interpersonal skills that are “didactic only.” The point is to avoid templated assignments such as “write a business plan,” which leave “no creativity, no building, and no agency.”
• Scaffolding without control: pods of roughly eight students where each week students report what they did the previous week and what they plan to do next, with a mentor present to push them. Biyani’s justification is that this mirrors adulthood: “once you enter the real world… there is nobody telling you you have to do anything.”
• He is explicit that none of this is novel — “Oxford still operates under it to this day,” alongside inspiration from YC, a16z SPEEDRUN, Montessori, and Alpha School, which he calls “probably one of the best high schools in the country.”

8. The founder’s own scar tissue: a 13-year-old builder that college stifled

• The origin story: Biyani’s first venture at 13 was technically a nonprofit that generated about $10,000 over a summer, with proceeds going to his high school. “The school was mostly trying to shut it down the entire time… my mother is probably the only person in my entire life who was like, ‘Gagan, you should keep doing this.’” He did not know startups were a career, got an economics degree, and has been reliving “what I wish I had when I was 13” ever since.
• His credentials-versus-building tension: a 1490 SAT, fives on nearly every AP, and a top public high school — yet rejected everywhere but the UC system because building cost him the 4.0. Berkeley “was stifling… I had seven jobs in three years… but I knew no one doing it. So I did it poorly.”
• The pipeline prediction: Alpha School students now starting at Berkeley and Stanford may find the transition from a high-agency environment difficult — “I am going to be tracking this like a hawk.” Beyond them, “there’s kids all over the country… who are misfits at their schools,” and Biyani thinks many parents understand that a stifling environment can limit their child’s upside.

9. Failure sucks, positive-sum culture, and why zigzaggers are bulletproof

• On teaching failure, Biyani refuses the romanticism: “I don’t believe in this mentality of making failure cool. Failure is not cool. Failure sucks… And yet if you fail, it’s okay. Get up and try again.” His proof is autobiographical: the Udemy co-founder falling-out, Sprig raising “a ton of money” and failing, and Maven succeeding, with his VP of product being promoted to CEO. Investors who lost money on Sprig still backed Maven and a16z Academy.
• The cultural claim, delivered as a career-security thesis: “the people who have the most job security are those who zigzag,” because classic consulting and investment-banking tracks “are getting blown up all the time,” whereas “if you have high agency and you can build, you’re essentially bulletproof.” The academy aims to create Silicon Valley’s “positive-sum culture”: “if you’re on top today, you may not be on top tomorrow. So you better be nice to everybody.”
• The closing book pick doubles as worldview: Hannah Ritchie’s Not the End of the World, found via Bill Gates — nuance “does not go viral,” but holding “climate change is a very serious problem” alongside the view that we are on track for it not to end the world is his template for AI. His memorable example: “You ever heard about the ozone in the last 15 years? You know why? Because we solved it. We fixed the ozone layer, Molly.”