As Silicon Valley races toward a future where AI agents do most of the software programming, a new problem is created: finding the AI-generated bugs before they are put into production. Even OpenAI is dealing with such issues, a former employee has described.
Newly funded startup PlayerZero has created a solution: use AI agents trained to find and fix problems before the code is put into production, the startup’s CEO and sole founder, Animesh Koratana, tells TechCrunch.
Koratana created PlayerZero while he was at the Stanford DAWN lab for machine learning under his adviser and lab founder, Matei Zaharia. Zaharia is, of course, a famed developer and the co-founder of Databricks; he created its foundational technology while working on his own doctorate.
PlayerZero on Wednesday announced that it raised a $15 million Series A led by Foundation Capital’s Ashu Garg, an early Databricks backer. This follows a $5 million seed led by Green Bay Ventures and several noteworthy angels, including Zaharia, Dropbox CEO Drew Houston, Figma CEO Dylan Field, and Vercel CEO Guillermo Rauch.
During his time at Stanford DAWN, Koratana, now 26, was working on AI model compression technology and “got exposed to language models really early on,” he says. He met the developers who crafted some of the first AI coding assistance tools.
It hit him then that “there’s this world in which computers are going to write the code. It’s not going to be humans anymore,” Koratana tells TechCrunch. ”What’s the world going to look like at that point?”
He knew before the term “AI slop” was even coined that these agents were going to produce code that broke things just as their human overseers do.
That problem would also be exacerbated by so many agents cranking out so much more code than has ever been written before. It won’t always be practical for humans to check all AI-written code for bugs or hallucinations. And the issue becomes even more intense for the large, complex code bases that enterprises rely upon.
PlayerZero trains models “that really deeply understand code bases, and we understand the way they’re built, the way they’re architected,” Koratana says.
His tech studies the history of an enterprise’s bugs, issues, and solutions. When something breaks, his product can then “figure out why and fix it, and then learn from those mistakes to prevent them from ever happening again,” Koratana says. He likens his product to an immune system for large code bases.
Landing Zaharia, his adviser, as an angel was a first step to fundraising, but the moment that really validated his idea was when he showed a demo to another famous developer: Rauch. Rauch is the founder of triple-unicorn developer tool company Vercel and creator of the popular open source JavaScript framework Next.js.
Rauch watched Koratana’s demo with interest but skepticism, asking how much of it was “real.” Koratana replied that this was code “running in production. Like, this is a real instance. And he was quiet,” Koratana says. Then his soon-to-be-angel investor responded, “If you can actually solve this the way that you’re imagining, it’s a really big deal.”
Of course, PlayerZero isn’t alone in attempting to solve the AI-generated bug problem. Just last week, Anysphere’s Cursor launched Bugbot to detect coding errors, as just one example.
Still, PlayerZero is already gaining traction for its emphasis on large codebases. While it was conceived for a world where agents are the coders, it is currently being used by several large enterprises that use coding co-pilots. For instance, subscription billing company Zuora is one of the startup’s marquee customers. Zuora is using the tech across its engineering teams, including to watchdog its most precious code, its billing systems, it said.