# Transcript: Featured Session: A Conversation with Garry Tan **Date:** March 14, 2026 · 10:00 PM **Session:** [Featured Session: A Conversation with Garry Tan](/sessions/2026-03-14/pp1149920-featured-session-a-conversation-with-garry-tan) ## Summary Y Combinator CEO Garry Tan, interviewed by Benchmark's Bill Gurley, discussed how AI is fundamentally transforming startup building — from one person doing the work of 2,000 engineers to the evolving YC application process. The wide-ranging conversation covered open-source model competition, California's proposed unrealized gains tax, and Tan's vision for a post-scarcity economy. ## Topics `y combinator` · `ai coding` · `startup ecosystem` · `venture capital` · `open source ai` · `vibe coding` · `california politics` · `founder skills` ## Key Takeaways 1. One person with AI tools can now be 2,000x more productive than a Google engineer in 2009 — Tan personally runs 10 AI coding workers on 3 projects while holding his full-time YC job. 2. The age of the product manager/designer founder has arrived: the critical skill is directing AI workers across many tasks, not writing code yourself. 3. Open-source and open-weight models are essential to prevent frontier AI labs from becoming the next monopolistic big tech — competition must be preserved. 4. YC's magic comes from socially constructed reality: surrounding yourself with other ambitious founders creates a cultural transformation that can't be replicated through individual mentorship alone. 5. Tech people must engage politically to protect innovation ecosystems — reasonable taxation is fine, but unrealized gains taxes and ineffective government spending threaten the golden goose. ## Full Transcript Y Combinator President and CEO Garry Tan was interviewed by Benchmark Capital's Bill Gurley at SXSW. Gurley praised YC's miraculous impact on the startup ecosystem, noting that many attempts to recreate what YC has accomplished have failed to come close. Tan shared that he couldn't even afford a badge when he was at SXSW in 2009 as a struggling startup founder. On the YC application process, Tan said everything has changed. He cares less about what university someone attended or where they work — he wants to see potential. The traditional requirement for co-founders is evolving: while having a co-founder still helps because the journey is long and 'reality is socially constructed,' solo founders using AI can now do the work of entire teams. Tan noted that one person can now be 2,000 times more productive than a Google engineer was in 2009. Tan described his personal AI coding workflow: while holding down his full-time job managing a near-trillion-dollar portfolio, he runs ten AI coding workers on three different projects simultaneously during early mornings and evenings. He can produce 10,000 lines of code per day across multiple projects. He recently open-sourced his workflow called 'G Stack' which has received about 9,700 GitHub stars. He described it as 'Brian Chesky in a box' — an AI system that helps you think about ideal user experiences and work backwards to implementation. The conversation turned to the AI landscape. Tan said about 80-90% of YC companies are software, with 10-15% increasingly in defense-related hard tech. Founders are using a mix of models: OpenAI and Anthropic for sophisticated reasoning, Grok and Llama for low-latency applications. He emphasized the critical importance of competition and open-source models to prevent any single frontier lab from gaining monopolistic pricing power. Gurley drew a parallel to the internet era: in 1996-97, thirty percent of venture dollars went to Sun and Oracle for the best technology, but five years later it was all Linux and MySQL. The same pattern of innovation phase followed by optimization phase is playing out with AI. Tan agreed, warning that frontier model companies are likely to become the next big tech — trying to lock up distribution and extract gross profit, just as Google and Meta did with consumer businesses. On the controversial California unrealized gains tax proposal, Tan was passionate. A 5% one-time tax on unrealized gains would mean founders with paper valuations could owe massive taxes even if their company later goes to zero. He's had friends leave California before January 1st as a precaution. Tan emphasized he's not against taxes — Massachusetts has a reasonable millionaire tax — but against ineffective government spending, citing million-dollar housing condos used as drug dens and the illegality of funding dry permanent supportive housing in California. Gurley asked about the types of companies that will emerge. Tan predicted this is the age of the product manager/designer founder rather than the pure engineer. The ability to direct AI workers across many tasks — like a CEO who has done every job in the company — becomes the critical skill. The shape of what society needs has changed: be 'a mile wide and a mile deep' in your core area, then 'an inch deep for the rest of the mile,' and let AI agents fill in everything else. On staying private versus going public, Tan acknowledged that public markets don't always support ambitious founders, making staying private rational. But he'd love to see more companies go public. Regarding YC expansion, Tan mentioned Cambridge, Massachusetts as a likely second location, especially given California's political challenges. The rate of unicorns is 2.5x higher in the San Francisco Bay Area, which keeps YC there despite the difficulties. Tan closed with his vision: technology should create a post-scarcity economy, solving real human problems like cancer through knowledge and understanding. He urged tech people to engage politically and speak up, sharing his own background of growing up food-insecure and relying on public schools. His latest project, 'There Is This,' is like 'WordPress and Meetup had an AI baby' — focused on building software for human connection, meaning, and community at 2,000x the current scale. --- *Source: stt · Language: en · Model: claude-opus-4-6* [← Back to session](/sessions/2026-03-14/pp1149920-featured-session-a-conversation-with-garry-tan)