Y Combinator expands AI agent use as Garry Tan says token spending reaches $200,000 a year

Traders Union
All News
Business & Wealth
Y Combinator AI agent spending

Highlights

Garry Tan reports personal annual AI agent token spending of $200,000, supporting work projects and coding tasks with extensive data integration.
Y Combinator deploys its internal agent, Stray Light, to overhaul operations by leveraging historic founder data, emails, and startup applications.
Tan highlights cost as a major barrier to mainstream AI agent adoption and expresses concerns about user autonomy amid dominance by large tech platforms like Meta.

AI agent spending and internal deployment

As reported by TechCrunch, Tan outlined the scale of his personal AI spending in an episode of Walter Isaacson's podcast The Builders, which is set to publish on Tuesday. He says he uses AI agents across work projects and coding tasks, and has also shared a set of 23 Claude Code skills, called gstack, on GitHub to show others what the tools can do.

Tan says his use of agents goes beyond standard research support. In addition to helping him prepare for interviews and questions, he says the systems have access to his email, Dropbox files, Apple Notes and even therapy notes dating back 20 years, allowing them to assist with a wide range of day-to-day tasks.

He also says Y Combinator is applying the technology internally through an agent called Stray Light. According to Tan, the tool is helping rebuild how the accelerator operates by drawing on historical context from founders it has backed, inbound emails and startup applications.

Cost barriers and control concerns

Tan argues that AI agents should not be treated like traditional software, saying excessive focus on repeatability and safety can limit the more human-like benefits these systems offer. He presents agents as tools that can act more autonomously, anticipate needs and help users make decisions more effectively.

At the same time, he says the economics remain a major obstacle to mainstream adoption. Tan describes his own spending level as a privilege and says personal AI agents will need to become much cheaper before they are widely accessible.

He is also cautious about a future dominated by free AI tools controlled by large technology platforms. Referring to concerns about algorithmic control at major social media companies such as Meta, Tan says he prefers a model in which users direct and program their own personal AI systems.

Our earlier article on a KPMG U.S. CEO survey examined how large companies are already seeing incremental efficiency and growth gains from AI agents, even though only a small share view them as transformational so far. It also highlighted that firms expect AI to drive continued hiring, retraining, and sustained investment in AI infrastructure as they redesign workflows around more autonomous systems.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.