Y Combinator CEO Garry Tan spends up to $200,000 a year on AI agent tokens

How much does Tan spend on AI agents?

Y Combinator CEO Garry Tan revealed on Walter Isaacson's podcast The Builders, set to publish November 4, that he spends roughly $100,000 to $200,000 a year on AI tokens, according to TechCrunch. The figure is roughly the cost of a senior engineer's salary.

Tan described his relationship with AI agents as intense, joking he has suffered from 'cyber psychosis' and sometimes barely sleeps because he is so excited to work with his agents. He uses them for both professional coding projects and personal life management.

As an early OpenClaw adopter who is now a power user, Tan says one thing he has learned is not to treat AI agents like traditional software. He told Isaacson that people overly focused on 'repeatability and safety' risk missing the 'beautiful parts of working with an AI agent, which basically can be so similar to a human.'

What does Tan use AI agents for?

Tan's AI agent has read all of his email, his Dropbox, his Apple Notes, and even his therapy notes going back 20 years, he told Isaacson. He describes the experience as having 'a Jiminy Cricket on your shoulder at all times' that knows you and proactively helps.

He has shared his personal toolkit of 23 Claude Code skills, called gstack, on GitHub to help others learn what agents can do. The toolkit is designed to be a teaching resource for developers getting started with autonomous AI coding agents.

Beyond coding, Tan uses agents to research podcast appearances and prep for interview questions. His agent also helps him run his life, proactively suggesting actions and helping him 'see around corners.'

What is gstack and why does Tan share it?

Tan's gstack repository on GitHub contains 23 Claude Code skills he uses daily, from code generation and refactoring to research and personal productivity. He open-sourced the toolkit to demystify what AI agents can actually do for developers and non-developers alike.

His philosophy is that agents should be treated more like collaborators than deterministic tools. The gstack repository includes configuration files and prompts that Tan has refined through months of heavy agent usage, providing a practical starting point for others exploring agentic workflows.

Y Combinator's own AI agent: Stray Light

Tan revealed that Y Combinator itself runs an internal AI agent named Stray Light. The agent has what he calls a 'company brain' with the total context of every founder YC has ever backed, anyone who has ever emailed the accelerator, and all applications submitted.

'We now have an internal AI agent, for instance, named Stray Light that kind of has a personality, and then it has a company brain,' Tan said, according to TechCrunch. The agent is 'rebuilding the way we operate at YC.'

What does Tan's spending say about the AI agent market?

Tan's personal spending of $100,000 to $200,000 a year on tokens illustrates both the promise and the affordability gap in the current AI agent market. While power users like Tan extract enormous value from agents, the price point puts serious agent usage out of reach for most individuals.

The revelation comes as the broader AI industry grapples with the economics of agentic AI. Elon Musk confirmed just days earlier that Grok Bot's usage limits were throttled because compute costs were 'unsustainably high relative to revenue,' highlighting that even well-funded AI operations struggle with token economics.

Tan's solution is not cheaper Big Tech-controlled agents but user-owned personal AI. 'You could imagine a personal AI that you control, that you program,' he said, pushing back against the platform model where companies control the algorithms users interact with.

'It's too expensive right now,' Tan said, adding that the amount he spends on AI agents is 'a crazy privilege that not everyone can do.'