Y Combinator CEO Garry Tan spends up to $200,000 a year on AI token costs to power his extensive daily use of autonomous agents and coding tasks. Appearing on Walter Isaacson’s podcast The Builders, Tan revealed that his heavy reliance on AI includes sharing his 23 Claude Code skills, called gstack, on GitHub to help others learn agent capabilities. His personal agents assist with work projects, research podcast appearances, and manage various aspects of his life by reading his emails, Dropbox files, Apple Notes, and decades of therapy records.
Tan describes the experience of working with autonomous agents as having a constant digital companion that helps him see around corners and run ahead on tasks. Beyond personal productivity, Y Combinator is utilizing an internal AI agent named Stray Light, which features a distinct personality and possesses the total contextual brain of every founder and application the accelerator has ever backed. While Tan believes personal AI agents will eventually become mainstream, he noted that current high costs represent a major barrier preventing widespread accessibility.
Reflecting on the broader technological landscape, Tan also expressed a preference for user-controlled personal artificial intelligence over free algorithms governed by massive tech corporations like Meta. He envisions a future where individuals can program and completely own their personal AI assistants rather than relying on centralized platforms. This high level of personal integration highlights the emerging frontier of autonomous agents in executive workflows and operational management.
Why it matters: Garry Tan's massive spending on AI tokens highlights how heavy power users are pushing autonomous agents to their limits while exposing cost barriers that must fal