Skip to content
worth noting AI agents

Y Combinator accelerator head Garry Tan described annual spending of up to 200 thousand dollars on AI agents

only one source so far

Garry Tan, head of the Y Combinator accelerator, said on The Builders podcast that he spends 100 to 200 thousand dollars a year on tokens for AI agents. He shares a set of 23 tools for Claude Code called gstack and mentioned YC's internal agent named Stray Light.

Garry Tan, CEO of the Y Combinator accelerator, said in an interview for Walter Isaacson's podcast The Builders that he spends roughly 100 to 200 thousand dollars a year on tokens for AI agents. He uses agents for work projects as well as programming and has publicly released his set of 23 skills for the Claude Code tool called gstack, which he uploaded to the GitHub platform in order, in his own words, to show others what agents can do.

According to Tan, his agent has access to all of his email correspondence, Dropbox, Apple Notes, and even therapy notes going back twenty years. He describes its role as a constant autonomous helper in everyday life that proposes further steps on its own. According to Tan, it's a mistake to treat AI agents like ordinary software — he says that excessive emphasis on repeatability and safety deprives users of the benefits an agent offers thanks to its similarity to a human collaborator.

According to Tan, Y Combinator has also deployed an internal agent named Stray Light, which has its own "company brain" with complete context on all founders who have ever applied for support from the accelerator, as well as everyone who has written to the accelerator.

Tan believes personal AI agents will become mainstream once the costs of running them come down — he himself acknowledges that his spending is "an enormous luxury" that not everyone can afford. At the same time, he prefers personal agents fully under the user's control over free variants from big tech companies, because he believes those would once again centralize control in a way similar to social networks.

What changed

Why it matters

The case illustrates the extreme end of the spectrum of AI agent use — from tens of thousands of dollars a month on tokens to an agent's access to sensitive personal data, including therapy records. At the same time, it documents a concrete direction in which companies like Y Combinator are building internal agents with complete context on clients and applicants to support operational decisions.

Two audiences, two different impacts

What this means

01

For individuals

Shows a real-world case of deploying a personal AI agent with access to email, cloud storage, and sensitive personal records, as well as a publicly available set of 23 tools for Claude Code (gstack) as a concrete source of inspiration for one's own workflow.

What to do Check out the publicly available gstack toolkit on GitHub as inspiration for working with Claude Code.
More practical updates →
02

For a business

Y Combinator developed an internal agent called Stray Light with full context on all founders and applicants, which is a concrete example of deploying a corporate knowledge agent for internal operations and decision-making.

Productivity
What to decide Consider whether an internal agent with context about clients or applicants similar to Stray Light would bring benefits to one's own operations.
More business impacts →
AI agents Claude Code Garry Tan gstack The Builders Y Combinator

Check the original

Event sources

only one source so far · 1 publisher, 1 independent. We count feeds from the same owner only once.

1
TechCrunch AI independent context · first detected Y Combinator’s Garry Tan uses AI agents so much, he spends up to $200K a year on tokens