A quarter of the W25 startup batch have 95% of their codebases generated by AI, YC managing partner Jared Friedman said during a conversation posted on YouTube.
…
In a video titled “Vibe Coding is the Future”, Friedman, along with YC CEO Garry Tan, managing partner Harj Taggar, and general partner Diana Hu, discussed the trend of using natural language and instincts to create code.
an important caveat to this, I think, is that YC is heavily invested in startups that will sell AI, not just startups that are using it to build their product. so they have an incentive to hype it up as much as they can.
if any of these startups succeed, my condolences to the engineers who get hired afterwards and are stuck bugfixing and trying to understand the LLM-generated codebase the founders slapped together.
This is any successful startup - you don’t succeed by making a perfect product, you succeed by making a buggy mess that’s enough to convince both investors and more importantly customers that there’s potential… That means you need to rebuild from scratch in years 2-4 anyway, so frankly for the engineers who are coming in then, there’s little to no difference
I think there is a difference. Because software is so flexible and quick to build, it’s orders of magnitude easier to build something known and understood.
A promising startup with its systems in a knot, but their initial team is still on retainer? Brains can be picked, abstraction boundaries placed, surgical rewrites deployed. Despite the mess, they still understand it, and development can expand.
It remains to be seen if AI-generated code is recoverable, if any existing strategies can be applied so humans can contribute, or if the company is forever beholden to AI providers to release a better AI to manage/improve what they’ve already got.