How to build custom AI agents the Y Combinator way

This article summarizes an episode of The Peel’s video series featuring Garry Tan, president and CEO of Y Combinator.

Garry Tan, president and CEO of Y Combinator / Photo credit: Garry Tan

At Y Combinator, the advice given to founders is evolving alongside technology. While the firm continues to fund the next generation of startups, its leadership is identifying a new competitive advantage: the ability to build and refine AI agents.

Garry Tan, Y Combinator’s president and CEO, views these tools as a way to change how things get built. He argues that automated workflows allow anyone to turn ideas into working products without the engineering hurdles.

The value of automating creative work

Trying to do all creative work manually is slow and repetitive. By choosing to build AI agents, creators can automate the most boring parts of their day.

“You can take almost anything that you might do all the time, and you just drop it into a context window,” Tan says. “[And tell it,] ‘Write me a prompt that can act as an agent that does…take this input and output over here.'”

A continuous feedback loop frees creators from starting over, letting them focus on refining their work

“Simultaneously, you start with the prompt, then you have an output,” adds Tan. “[Then you say,] ‘Give me the next version of the prompt…based on this conversation, give a better prompt that incorporates all the things we talked about.’…My YouTube prompt one is on version 27 or something.”

Unlocking billions of new builders

Letting AI handle the repetitive work makes creators more efficient and opens up a whole new way to build software, it’s no longer just for engineers.

“There’s a tiny percentage of GDP that is touched by software at all,” Tan explains. “There are only 10 to 20 million people in the world who are actually good at code. We’re about to enter this other moment where billions of people can actually just very directly create their own stuff.”

Lowering the barrier to entry requires new demand, but Tan argues that making things faster and cheaper can generate new opportunities.

While some worry about the economic impact of AI, he dismisses this fear. Since new tools naturally spawn new ecosystems, he sees massive potential.

“That’s what’s going to happen. A lot of things get better, faster, cheaper. We need more markets,” he points out. “[Without them,] the result is like all kinds of just loss…and then a market was created out of thin air, and then a lot more of it [business] is done.”

Turning ideas into AI workflows

Using this new creative power means dealing with practical implementation. To succeed, builders have to follow a structured system to turn their ideas into a functioning AI workflow.

Rethinking the value of content

Defining the creative baseline