Technology & Hacking · VIDEO
Garry Tan spends $200,000 a year on tokens for his AI agents: here's how he uses them
Garry Tan, CEO of Y Combinator, revealed that he spends between $100,000 and $200,000 a year on tokens for his AI agents, using them for both work and personal tasks. The agents, such as Stray Light, handle sensitive data and business operations, but the high cost remains an obstacle to mainstream adoption.
KEY TAKEAWAYS
- Garry Tan spends between $100,000 and $200,000 annually on tokens for his AI agents.
- The AI agents handle emails, personal notes, and business operations like Y Combinator.
- Tan emphasizes the importance of not treating AI agents as traditional software.
- The high cost of AI agents is the main obstacle to mainstream adoption.
- Custom AI agents could be the solution to Big Tech's algorithmic controls.
Transcript
Garry Tan, CEO of Y Combinator, has revealed he spends between €100,000 and €200,000 annually on tokens to power his AI agents. These advanced tools manage his work and personal tasks with unprecedented autonomy.
Tan shared this information on Walter Isaacson's podcast "The Builders," describing how these agents handle emails, personal notes, and documents dating back decades. Y Combinator uses an AI agent called "Stray Light," which has developed a distinct personality and comprehensive knowledge of the organization's communications.
The main obstacle to widespread adoption of these AI agents is their high cost. Tan advocates for personal AI agents over centralized solutions, promoting his set of 23 AI skills called gstack, available on GitHub. He believes AI agents should operate with a degree of autonomy, similar to a personal assistant, requiring a cultural shift in technology interaction.
Despite the benefits, Tan acknowledges that the current cost is a significant barrier. He suggests that for AI agents to become mainstream, they need to become significantly cheaper or find alternative business models.
The extensive use of AI agents also raises important issues of cybersecurity and data management. Organizations will need robust solutions for data loss prevention and security management to ensure these tools do not become points of vulnerability.
As AI agents become more widespread, the need for an adequate regulatory framework will emerge. This includes considerations on NIS2 compliance and digital operational resilience.
For accelerators like Y Combinator, adopting internal AI agents could revolutionize startup support. However, integrating such technologies requires a strategic approach to maintain the human relationship between accelerators and startups.
The AI agent market is rapidly evolving, with companies like Microsoft and Google developing centralized solutions, while startups focus on more personalized approaches. The fundamental difference concerns data control and operation.
