Y Combinator CEO Garry Tan has jokingly admitted in the past that he suffers from “cyber psychosis” and is sometimes so excited about working with AI agents that he can barely sleep. In a new episode of Walter Isaacson’s podcast “The Builders,” scheduled for release Tuesday, Tan puts a price tag on his passion.
“I probably spend about $100,000 to $200,000 a year on tokens,” he told Isaacson. He uses the agent for countless work projects and coding tasks, and freely shares his 23-piece code skill set, called gstack, on GitHub. He said this is an effort to help teach others what agents can do.
In the meantime, he uses the agent in the same way many people use AI, studying Isaacson’s podcast to help prepare him for the kinds of questions he might be asked. On the other hand, his agent can also help him make a living.
“All of my emails are being read, all of my Dropbox is being read. For example, all of my Apple notes are being read, even my therapy notes from 20 years ago are being read,” he said.
He describes the use of AI agents as “helping you with almost everything you do in your daily life. AI agents can do that more autonomously. Think of it like Jiminy Cricket, always on your shoulder. Like, pretty much everything you want to do, the AI knows you and can help you. And it can actually proactively say, ‘How about this?’ Or it can help you see around the corner. ”
As an early adopter of OpenClaw and now a power user, Tan says he has learned not to treat AI agents like traditional software. Some people “get so worried about reproducibility and safety that they end up sacrificing all the great things about working with AI agents that are basically very human-like,” he said.
For example, Y Combinator also uses agents who are “reinventing how we operate at YC,” he said. “For example, we currently have an internal AI agent named Stray Light, which has a kind of personality and also has the brains of the company. It’s like we actually have the full context of every founder we’ve ever supported, every person we’ve ever emailed, every application.”
Naturally, Tan, a technology enthusiast (as does anyone who runs Silicon Valley’s famous startup accelerator), believes that personal AI agents will eventually become mainstream once one big problem is solved: They need to come down in cost.
“It’s too expensive right now,” he said, adding that he knows spending money on AI agents is “an incredible privilege that not everyone has.”
But Big Tech-controlled giveaway agents probably aren’t the answer either. When discussing the future we want technology to create and how Meta is impacting us as a social media company, he said: “This is a huge company that controls algorithms, and we all hope on some level that they don’t.” “So imagine a personal AI that you control and program.”
If you make a purchase through links in our articles, we may earn a small commission. This does not affect editorial independence.
