Y Combinator CEO Garry Tan Spends Up to $200K a Year on AI Tokens

Garry Tan, the chief executive of Y Combinator, disclosed on Walter Isaacson’s podcast The Builders that he allocates between $100,000 and $200,000 annually to AI token usage. The revelation highlights how deeply personal AI agents have entered the workflow of a leading Silicon Valley investor and raises questions about the scalability of such tools for the broader founder community.

Tan described his agents as “a Jiminy Cricket on your shoulder,” noting that they read his email, Dropbox files, Apple Notes and even two decades of therapy notes. He relies on the agents for research, coding, and day-to-day decision making. To demystify the technology, Tan published a collection of 23 Claude Code skills, branded as gstack, on GitHub, inviting other developers to explore the capabilities of AI-driven assistants.

Within Y Combinator itself, an internal AI agent named Stray Light is being used to reshape the organization’s operations. Tan explained that Stray Light possesses a personality and a “company brain,” giving it access to the full history of every founder backed by YC, all email correspondence, and every application the accelerator has received. The tool is intended to surface insights and automate tasks that would otherwise require manual review.

Despite the productivity gains, Tan cautioned that the current price of AI tokens is prohibitive for most entrepreneurs. He called his own spending “a crazy privilege” and emphasized that widespread adoption will depend on a significant reduction in cost. Tan also voiced skepticism about relying on free agents supplied by large platforms, arguing that a personal AI that users can program and control would better serve individual needs without ceding algorithmic power to a single corporation.

For founders watching the conversation, Tan’s experience serves as both a proof of concept and a warning. The ability of AI agents to ingest massive personal data and provide proactive suggestions could transform how startups operate, yet the financial barrier may limit early-stage companies to lighter-weight tools. As token pricing evolves, the industry will likely see a push toward more affordable, open-source alternatives that retain the autonomy Tan envisions. The next wave of AI-enhanced productivity will hinge on balancing capability, cost, and control for the broader entrepreneurial ecosystem.