Analysis of Y Combinator’s Garry Tan uses AI agents so much, he spends up to $200K a year on tokens - aVenture Research
TechCrunchby Julie Bort
Y Combinator’s Garry Tan uses AI agents so much, he spends up to $200K a year on tokens
Y Combinator Garry Tan has, in the past, jokingly admitted that he's suffered from "cyber psychosis” and barely sleeps sometimes because he’s so excited to be working with his AI agents.
Analysis
These are Tan’s descriptions and expenditure estimate, not audited company spending or measured productivity outcomes.
The specific business significance is a concrete signal of intensive demand for agentic inference alongside a sharply stated affordability constraint: Tan says personal agents are too expensive for most people at present.
Together these findings make falling unit prices insufficient evidence that full agent workloads are becoming cheaper or that providers will capture healthy margins.
Whether lower prices and better efficiency unlock enough valuable usage to offset greater workflow complexity remains unresolved; the article supplies no cost-per-task, output-quality, or productivity comparison for Tan’s own usage or YC’s internal agent.
Tan’s role is that of an unusually intensive user and public advocate of agents; he says the level of spend is a privilege, while his YC example places the technology inside an organization’s handling of founder and applicant context. For Y Combinator, this is an operational exposure as well as a demonstration: the value proposition depends on whether broader context and agent autonomy improve work enough to justify inference expense and the risks of consolidating sensitive correspondence and applications in an internal system.
The article reports Tan’s account, not independently measured savings, effectiveness, or safeguards, so it does not establish that this deployment is replicable at typical startup or accelerator budgets.