Descriptions:
a16z sits down with Y Combinator president Garry Tan for a wide-ranging conversation that spans founder psychology, Silicon Valley culture, and how AI is reshaping what it means to build a startup in 2026.
Tan uses his own career arc — joining Microsoft’s Windows Mobile team after the dot-com bust instead of staying with web programming — as a framework for arguing against trend-chasing and toward earnestness and direct experience. The conversation explores whether high-agency is innate or developed, how YC coaches founders to run business feedback loops (not just product loops), and what distinguishes startups that can move on AI from large organizations that can’t.
On the AI building side, Tan shares his own experiments: GarysList.org as an agentic newsroom, and a RAG-powered personal research tool called Soda that he says outperformed Memory Palace and other open-source alternatives in head-to-head testing. He argues the next competitive edge for AI-native startups lies in context management — specifically, knowing which tokens to keep in an agent’s working memory for a given task — and that frontier model labs’ compute constraints create an opening for builders who architect around those limits. His broader point is that the same contrarian instinct that led early founders to bet on the web during the dot-com bust applies today to founders building deeply with AI before the mainstream catches up.
📺 Source: a16z · Published August 12, 2026
🏷️ Format: Interview
