What Garry Tan said about spending and agents
Garry Tan AI agents sit at the centre of how the Y Combinator chief executive works, and on Walter Isaacson’s podcast The Builders he put a rough price on that habit. Tan said he spends about $100,000 to $200,000 a year on tokens—the metered units many AI systems bill for—while using agents across coding, research and day-to-day tasks. He has also joked in the past about “cyber psychosis” and barely sleeping at times because he is so excited to work with those agents.
An artificial intelligence agent, in the sense Tan describes, is software that can take on many ordinary tasks with more autonomy than a simple chatbot: he likened it to “a Jiminy Cricket on your shoulder at all times,” something that knows you, helps you see around corners, and can run ahead with suggestions.
How he uses agents at work and in daily life
Tan said he uses agents the way many people use AI—to research a podcast guest and prepare for likely questions—and also to help run his life. “It’s read all of my email and all of my Dropbox. It’s read, like, all of my Apple Notes, even my therapy notes going back 20 years,” he told Isaacson.
He has freely shared a set of 23 Claude Code skills, called gstack, on GitHub, saying the aim is to help teach others what agents can do. As an early OpenClaw adopter who is now a power user, he argued that people should not treat AI agents like traditional software. Worrying too much about “repeatability and safety,” he said, can mean trading away “the beautiful parts of working with an AI agent, which basically can be so similar to a human.”
Inside Y Combinator and the cost barrier
Garry Tan AI agents are not only a personal toolkit. He said Y Combinator is using an agent that is “rebuilding the way we operate at YC,” including an internal agent named Stray Light that “kind of has a personality” and a “company brain,” with context on every founder the accelerator has backed, people who have emailed the firm, and applications.
He believes personal AI agents will become mainstream only if costs fall. “It’s too expensive right now,” he said, calling the money he spends on agents “a crazy privilege that not everyone can do.” Free agents controlled by large technology firms are not his preferred end state either. Pointing to how a major social platform controls its algorithm, he imagined instead “a personal AI that you control, that you program.”
Aasan alfaaz mein
Y Combinator ke CEO Garry Tan ne kaha ke woh AI agents ke liye saal mein lagbhag $100,000 se $200,000 tokens par kharch karte hain. Yeh agents unke kaam, coding aur rozmarra ki zindagi mein madad karte hain, email aur notes tak. Unhone kaha ke abhi yeh bohot mehnga hai, is liye aam logon ke liye asani se mumkin nahi. Jab qeematain kam hongi, tab personal AI agents zyada logon tak pohanch sakte hain.
