Garry Tan: "Personal AGI Is How You Stay Under Your Own Power"
YC President Garry Tan introduces "Personal AGI," a concept where individuals own and train AI agents on their unique context, transforming personal productivity and startup creation. He shares his workflow, advocating for Markdown as code and emphasizing the critical importance of owning one's intellectual assets in the age of AI. The talk outlines a practical 5-step guide to building a personal AI system and explores the profound implications for innovation and individual empowerment.
Garry Tan, President & CEO of Y Combinator, outlines a vision for “Personal AGI” – a paradigm where individuals harness AI agents running on their owned infrastructure to dramatically amplify their capabilities and build with unprecedented autonomy. He draws inspiration from the philosopher Spinoza, who, despite being ostracized, pursued truth and built tools independently, embodying conatus (striving).
What is Personal AGI?
Tan defines Personal AGI as general intelligence tailored for one person, you. It’s an agent that runs on your infrastructure, reads from a memory you own, executes procedures you wrote, and continuously compounds your knowledge. This is explicitly contrasted with “corporate AGI,” which are rented, subscription-based chatbots that reset and are controlled by external companies.
The Productivity Multiplier
Tan asserts that Personal AGI offers an exponential increase in productivity. He estimates his own output has increased 400x (conservatively 8x) since 2013, a phenomenon he observes across YC’s portfolio, where small teams are achieving historically high revenue per person. This multiplier applies to all knowledge work, not just coding. The key leverage point is not the frontier models (which are commodities) but the context you feed them and the harness you use to orchestrate them.
Your Life as a Library
Human working memory is severely limited (7 +/- 2 items). AI agents, however, can process vast amounts of information (millions of tokens). The crucial insight is that your life—your emails, meetings, notes, decisions—is a library. A Personal AGI needs not just a library but a librarian to intelligently access and synthesize this personal context. Tan’s system, GBrain, is a knowledge wiki containing 25 years of his digitized life, curated and searched by agents, enabling instant recall of relevant information.
Inside Garry Tan’s Personal System
Tan details his practical implementation:
- GBrain: A personal knowledge wiki with ~220,000 Markdown pages of his life data.
- Daily Workflow: Agents process his inbox, triage emails, provide context-rich briefings, generate meeting prep docs, and conduct research overnight.
- GStack: An open-source agent coding framework built on GBrain, enabling agents to interact with the world via “skill files.”
Markdown is Code & The Latent/Deterministic Divide
Tan argues that Markdown is code: clear English instructions can be executed by a language model, effectively making anyone who can write clear instructions a programmer. This empowers non-technical individuals to build sophisticated automations.
He also stresses the importance of understanding where computation should occur:
- Latent Space: For judgment, taste, and interpreting vague requests (handled by the model itself).
- Deterministic Space: For precise, arithmetic, or data-driven tasks (handled by external code, databases, or scripts). Effective Personal AGI combines both, with Markdown skill files calling deterministic code when needed.
Building a Company of One
Personal AGI changes the fundamental “physics” of startups, enabling individuals to create and manage an entire “organization” of agents before even forming a company. This allows founders to execute