Aug 11, 2026 · 51m · a16z
Garry Tan: New Rules for Founders
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of The a16z Show, Y Combinator President and CEO Garry Tan joins host Anish Acharya to discuss the evolving mechanics of tech startups, the transformative impact of AI agents on founder productivity, and the crucial role of earnestness, high agency, and civic leadership.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. How this is scored →
The host as informed peer 4.8 Guest teaching 2.3 Guest disagreement 1.5 The host pushing back 1.3
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 47:45 Condemning institutional failure and media silence
Garry forcefully criticizes SF governance, ideological school boards, and mainstream newsrooms for failing to protect vulnerable communities.
Hardest push from the host ▶ 20:12 Challenging whether agency can actually be taughtAnish directly challenges Garry's assertion that agency is malleable, quoting Nikita's thesis that only a static percentage naturally produces creative output.
Biggest teaching moment ▶ 8:04 The $4B lesson on following the map instead of territoryGarry provides a masterclass on founder instinct by breaking down his multi-billion dollar blunder of choosing a corporate promotion over Palantir.
The host holds their own ▶ 42:22 Analyzing token pricing economics in consumer softwareAnish displays high-level domain mastery by articulating how collapsing marginal token costs are the requisite unlock for true consumer AI software.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The host as informed peer | Guest teaching | Guest disagreement | The host pushing back | Why |
|---|---|---|---|---|---|---|
| Welcome to The a16z Show with Garry Tan | 4 | 2 | 1 | 1 | Anish warmly introduces Garry and frames the discussion around founder psychology versus Silicon Valley history. Garry reflects collaboratively on his early career choices post-dot-com crash and the mistake of chasing trends rather than core knowledge. | |
| Founder Courage and First-Principles Curiosity | 5 | 2 | 1 | 2 | Anish synthesizes Garry's point on earnestness being grounded in courage and references Chris Dixon's idea maze concepts. Garry agrees warmly and elaborates on the punk/outsider origins of Silicon Valley movements. | |
| The Palantir Regret and Cult-Like First Principles | 4 | 3 | 2 | 1 | Anish asks about the shift in Palantir's status over time, and Garry shares his multi-billion dollar regret of turning down Peter Thiel's early offer for a promotion at Microsoft. Garry explains how cult-like conviction and first-principles gnosis define great companies. | |
| How Y Combinator Democratizes Silicon Valley | 4 | 2 | 1 | 1 | Anish prompts Garry on how YC systemically transforms outsiders into Silicon Valley insiders. Garry explains the structural model created by Paul Graham and Jessica Livingston to bypass traditional social gatekeepers. | |
| The Evolution of Founders: Vibe Coding and AI Agents | 5 | 3 | 2 | 1 | Garry outlines the shift from requiring co-founders to solo vibe-coding where one person wields the power of hundreds of agents. He boldly predicts the demise of pure per-seat SaaS models over the next 5-10 years. | |
| Code Is No Longer Precious: Agency and Taste | 5 | 2 | 1 | 1 | Anish notes how low code friction makes software creation less precious, and Garry discusses open-sourcing G-Stack and G-Brain tools. Garry emphasizes that taste and agency have become the supreme differentiators. | |
| Is Agency Innate? Garry's Late-Bloomer Journey | 6 | 3 | 2 | 4 | Anish explicitly challenges Garry's premise that agency can be learned by citing Nikita's counterargument about the persistent 20% creative minority. Garry defends agency as a practiced muscle by citing his own late-bloomer trajectory, before deep diving into agentic RAG and token-maxing architecture. | |
| Agent Provenance and the Rise of Senior Founders | 5 | 2 | 1 | 1 | Anish asks where agent recursive loops fail, prompting Garry to discuss provenance and why 35-45 year-old senior founders possess an operational advantage. Garry shares how Pedro from Brex uses CrabTrap and transcripts to resolve cross-functional conflict. | |
| Overcoming Cognitive Limits to Build a Better World | 4 | 3 | 2 | 1 | Garry explains human cognitive bottlenecks (7 plus or minus 2) compared to LLM context windows capable of digesting multiple books of state. He recounts pushing back against a WSJ reporter trying to trap him on tech utopianism. | |
| Bypassing Corporate Hell: The Microsoft Baseball Bat | 5 | 2 | 1 | 1 | Anish and Garry discuss the managerial class morass, prompting Garry to share an anecdote about bringing a baseball bat to the Windows team at Microsoft to resolve dependency blockers. Both agree startups must replace internal friction with cron loops. | |
| Re-architecting Companies: Lessons from Toyota | 5 | 2 | 1 | 1 | Anish proposes that middle management should be entirely probabilistic agents. Garry builds on this by referencing the Toyota Production System and showing how API lines disappear when frontline workers and agents continuously modify the pipeline. | |
| The White Pill: Why Societal Friction Prevents AI Doom | 4 | 2 | 2 | 1 | Garry delivers his 'white pill' contrarian thesis: organizational inertia and human slowness prevent sudden AI displacement doom, giving society an orderly 20-year transition window. | |
| The Next Form Factor: Voice, Memory, and Harness Wars | 6 | 2 | 1 | 1 | Anish and Garry explore future form factors, predicting voice interfaces and upcoming harness wars. Anish provides deep expertise on token pricing dynamics and the necessity of zero-marginal-cost distribution for consumer software. | |
| Civic Leadership: Protecting Communities and Local Governance | 5 | 2 | 3 | 1 | Garry passionately details his civic activism in San Francisco around middle school algebra, elder safety, and combating political complacency. Anish praises Garry's civic stewardship and extends the insight to local ambition globally. |
Statements from this episode (25)
Tan: Founders should avoid following the crowd when sectors are declared dead
“And then you shouldn't follow what the crowd is saying, right? Like, at that moment, people said the web was dead, but that was sort of the perfect time to work on web and social and things like that and, you know, I didn't do it.”
Garry Tan: Founders should pursue unique interests instead of asking what's hot
“The classic Question I'm sure you get that, you know, all investors get all the time is what's hot and what should I work on? And that's the wrong question to ask. Like the right question is like, what are you interested in? What do you know uniquely?”
Acharya: Truly interesting startup ideas lack a defined intellectual edge
“I think a good measure of how interesting the thing is, is can you find the edge? And for the most interesting things, you typically can't. Even if you find the edge of the conversation, you can't find the intellectual edge, which I think is a very good sign.”
Tan: Turning down early Palantir offer was a $2B to $4B mistake
“Peter said, I'm so sure this is the right thing for you. Like, here's a check for 70 grand. That was how much I made at Microsoft at the time. And I said, thank you very much, Mr. Thiel, but I might get promoted to level 60 this year. Which I did, but I think …”
Tan: Great ventures begin as cults with beliefs defying orthodoxy
“From Palantir, I definitely learned that everything that's awesome in my life is kind of a cult. That starts with some sort of truth or belief that flies in the face of an orthodoxy, which is very punk, actually.”
Tan: AI and vibe coding multiply developer productivity by hundreds
“What's happening right now with vibe coding and agentic coding literally any given person could be 400 of that person, like, from two years, like, even, like, nine months ago you can be a multiple of yourself.”
Tan: Pure per-seat SaaS may disappear within 5 to 10 years
“Like, you know, I was just doing an analysis on, like, pure sass, and I was like, oh, shoot, like, you know, a pure, like, per seat sass thing, like, not totally clear it will exist in another five or 10 years.”
Tan: YC advice prompt was reduced 90% in Claude before open-sourcing
“And then I actually had to go into Claude and I said, reduce this by 90%. Reduce the strength and intensity of this prompt by 90%. And then I open sourced that. Cause some of it's like, you gotta come and, you know, do YC.”
Tan: Agency and taste are off the charts important now
“No, I mean, and then I think it's crazy to what degree, like, suddenly agency and taste, like, off the charts important now.”
Tan: Operating AI agents at full strength costs $50K-$100K annually
“It costs, I don't know, 50 or a 100,000 dollars a year to, like, use the agents at full strength, like, full 150 IQ.”
Tan: A markdown file is an employee that executes tasks perfectly
“A markdown file is an employee. And it's an employee that will do the job perfectly every single time. And it'll do it like as many times as you want.”
Tan: Companies can reach $15M ARR in four months with two people
“Literally, you can go from zero to 15 mil ARR in about four months with, like, two or three people and, like, hundreds of agents. Like, you know, a few hundred skill files.”
Tan: Brex co-founder built open source layer Crab Trap to monitor agents
“What he ended up doing was, like, Building an entire open source layer, which he actually has open source called Crab Trap. That is like an agent that watches all the network traffic and all the things that Open Claw does.”
Tan: Brex CEO uses AI agents to analyze direct reports' meeting transcripts
“He has his agent look at the meeting transcripts from, like, All of his direct reports, and he, like, can see that, like, two levels down. And then the thing that really jumped out at me is, like, he can use meetings that he's not in. To figure out, like, what…”
Garry Tan: Existing institutions are constrained by human working memory limits
“And so the whole world, like all business, all things that people rely on, government especially is basically still built on a world that is predicated by human beings, limited human beings that can only keep seven plus or minus two things in their head.”
Tan: Startups use AI to compress six-month product releases into days
“Like you could release something in a day or two that would have taken like six months. And if you could like multiply that out across, like the thing is like probably an org like Microsoft can't. But, like, a startup can. And every startup must.”
Acharya: Organizational management handled by product managers should be automated
“Just the whole organization layer, which a lot of product managers historically did, should not be human.”
Tan: Y Combinator funded Clawvisor, which built an agent-only bug reporting API
“There's a, API company that we funded called Clawvisor, and they released a bug report endpoint that is not for humans to use. It's for agents who use their software to use”
Garry Tan: Full societal and corporate adoption of AI will take 20 years
“So I think that, you know, the white pill to me is like, it's gonna be 20 years, and that's not a bad thing. That's actually a good thing. Yes. Like, society is way slower than you think. Government is way slower than you think. Every company in the world is w…”
Garry Tan: AI startups will not replace incumbent giants like Microsoft
“There are real motes to these things. Like, there are structural reasons why a Microsoft isn't going anywhere. And then, I guess, This is sort of a little antithetical to what most, you know, most people in Silicon Valley would say, like, I want to start up to…”
Garry Tan: Next computing form factor will clearly be voice and memory
“I think in the short term it'll probably keep looking like the way it is, but it seems inconceivable that it'll stay that form factor. Like, it's pretty clearly going to be voice. I think there's a lot around memory.”
Tan: AI 'harness wars' for 1B consumers will erupt around 2027
“My guess is, like, next year, twenty-twenty-seven, will be, like, the harness wars. You know, and then, but it's really a cost thing. Like, it's just too expensive right now. You know but it won't be, like, there will be a war for a billion consumers. Right wh…”
Tan: San Francisco ABC affiliate refused to cover anti-Asian hate crimes
“ABC News you know, literally the mainstream media TV channel of San Francisco and the Bay Area would not cover anti-Chinese and anti-Asian crime.”
Tan: Fixing local governance will repair national politics within 10 years
“If we fix local, like, state and national will fix itself, because, like, On a five or 10 year basis, like the people who win up there, like, they will come up through a system that's functional and about saying no to things that don't make sense, about saying…”