'Vibe coding' lets 10 engineers do the work of a team of 50 to 100, says CEO of Silicon Valley incubator

• Y Combinator CEO Garry Tan said "vibe coding" is letting startups stay leaner.
• He said coding with AI allows smaller teams to do more heavy lifting in an interview with CNBC.
• He suggested young engineers struggling with the job market should "vibe code" and build startups.

The CEO of Silicon Valley's most famous incubator thinks "vibe coding" is set to transform the startup landscape.

"I mean, the wild thing is people are getting to a million dollars to 10 million dollars a year revenue with under 10 people, and that's really never happened before in early stage venture," Garry Tan, CEO and president of Y Combinator, said in an interview with CNBC.

Tan said it's due, in part, to vibe coding — Silicon Valley's favorite new buzzword, coined by OpenAI cofounder Andrej Karpathy in a post on X in February.

But what, exactly, does vibe coding mean?

"You can just talk to the large language models and they will code entire apps," Tan said. "And if it doesn't do — if there's a bug, or if you want it to change, or you want it to look a different way, you don't have to, you know, go in there and write the code yourself."

In essence — Tan seems to define it as an increasing reliance on artificial intelligence to do the bulk of the heavy lifting during the programming process. He said vibe coding is dialing up the speed at which startups can develop relevant software.

"You don't have to hire someone to do it, you just talk directly to the large language model that wrote it and it'll fix it for you," he said. "And sometimes you could just accept all changes without even looking at the changes it made, because it's that good now."

Tan says vibe coding makes the process of building software overall more efficient, and the current batch of startups that Y Combinator is incubating — which Tan says is composed of about "81%" AI companies themselves — is taking full advantage.

"This is the first time that's ever happened, and about 25% of the batch — 95% of their code was written by large language models," Tan said.

The sheer power of LLMs is allowing startups to stay leaner, Tan said. What would've once taken "50 or 100" engineers to build, he believes can now be accomplished by a team of 10, "when they are fully vibe coders."

"When they are actually really, really good at using the cutting edge tools for code gen today, like Cursor or Windsurf, they will literally do the work of 10 or 100 engineers in the course of a single day," he said.

Vibe coding does have its drawbacks, Tan caveated in an episode of Y Combinator's Lightcone Podcast earlier this month. In particular, a survey of Y Combinator's current batch of founders indicated LLMs are weak at actually debugging the code they churn out.

"The humans have to do the debugging, still. They have to figure out well, 'What is the code actually doing?'" he said, adding, "There doesn't seem to be a way to just tell it, 'debug.'"

Still, to Tan, the benefits are plentiful. Among them — that investing time and money into building niche software is more justifiable, given the new speed at which AI makes it possible to code.

"I think the other thing that it's going to do in the industry is that there are all these pieces of software that normally you could never write software for, like industries that before people would say, 'Oh it's too small a market,'" he said.

All of those narrower markets, he added, can now support business that gross a "hundred million" yearly, while still being run by small teams.

"This is really the good news," Tan said.

For those struggling to break into a shrinking job market, Tan says the advent of vibe coding has arrived at the "perfect time." It gives young engineers the opportunity to strike out on their own rather than relying on big-name companies to kickstart their careers.

"You know, maybe it's that engineer who couldn't get a job at Meta or Google, who actually can build a standalone business making 10 or 100 million dollars a year with 10 people," Tan said. "Like that's such a powerful moment in software."