A quarter of startups in Y Combinator's latest batch have codebases that are almost entirely AI-generated. That’s not a gimmick—it’s a shift in how software is built.
Jared Friedman, a YC managing partner, says these founders could code everything themselves. They’re technical enough. But instead, they’re leaning on AI to do the heavy lifting.
It’s an approach that’s been growing, thanks to "vibe coding"—a phrase coined by Andrej Karpathy from OpenAI. Instead of writing code line by line, developers just describe what they need, and AI handles the rest.
The Good: Speed and Accessibility
For early-stage startups, speed is everything. If AI can generate a functional MVP in days instead of months, that’s a massive advantage.
It also means more non-technical founders can launch software products without needing to hire developers right away.
Then there’s cost. Startups always have to be careful with money, and AI can cut down on expensive engineering hours.
The promise here is that founders can focus on business strategy instead of sweating over syntax errors.
The Bad: Messy Code and Security Risks
But AI-generated code isn’t perfect. It tends to be bloated, inefficient, and sometimes insecure.
Founders who rely too much on AI risk shipping products full of vulnerabilities.
Debugging is also a problem. If you don’t fully understand the code AI spits out, fixing it becomes a nightmare. YC’s CEO, Garry Tan, warns that solid coding skills are still necessary, especially as startups scale.
AI can generate code, but making sure it works in real-world conditions is still a human job.
The Future: AI as a Co-Pilot, Not the Pilot
This AI-driven approach to software development isn’t going away. As tools get better, startups will keep using them.
But the most successful founders won’t just rely on AI blindly—they’ll know when to step in and take control.
So, is AI-written code the future?
Yes, but only if founders know how to manage it. Otherwise, they might be building a house of cards—one bad deployment away from disaster.
