The talent wars in AI just dealt Mira Murati's startup a significant blow. Thinking Machines Lab is losing two co-founders, Barret Zoph and Luke Metz, who are headed back to OpenAI in a move that highlights the competitive pressure facing well-funded AI startups. The departures come less than a year after the company's $2 billion seed round valuation and underscore how difficult it is to hold onto top talent in the race to build advanced AI systems.
The exodus was announced in staggered fashion on Wednesday. Mira Murati, Thinking Machines Lab's CEO and a former CTO of OpenAI, first revealed the split in an X post, saying the company had "parted ways" with Barret Zoph. "Soumith Chintala will be the new CTO of Thinking Machines," she wrote. "He is a brilliant and seasoned leader who has made important contributions to the AI field for over a decade."
What Murati's announcement didn't mention was that the departures extended beyond Zoph. Just 58 minutes after her post, Fidji Simo, OpenAI's CEO of applications, dropped the full details on X: Zoph, Luke Metz, and Sam Schoenholz were all heading back to OpenAI. "This has been in the works for several weeks," Simo wrote, "and we're thrilled to have them join the team."
The coordinated timing raises questions about what really happened behind the scenes. According to Simo's account, the moves weren't sudden or emotional departures. They'd been in planning for weeks, which means Murati and Thinking Machines' leadership knew these departures were coming. That makes the staggered announcement strategy somewhat puzzling - Murati acknowledged Zoph's departure while remaining silent about Metz and Schoenholz, only for OpenAI to fill in the gaps minutes later.
Zoph brings serious AI credentials to OpenAI. He previously served as VP of research there and spent six years at Google as a research scientist. Metz, who co-founded Thinking Machines alongside Murati and Zoph, had spent years on OpenAI's technical staff. Schoenholz, whose LinkedIn profile still listed Thinking Machines as his current employer, had also come from OpenAI.
This is the second significant blow to Thinking Machines Lab in recent months. Co-founder Andrew Tulloch departed for Meta back in October, marking another loss of a founding-team member. While recruiting movements between AI companies are normal in Silicon Valley, losing two co-founders simultaneously - especially when one serves as CTO - signals deeper challenges at the startup.
Thinking Machines Lab has had a meteoric rise. The company raised $2 billion in its seed round last July, a staggering sum for a pre-product company that valued it at $12 billion. The round attracted top-tier investors including Andreessen Horowitz, Accel, Nvidia, AMD, and Jane Street. The capital infusion and valuation underscore how much the industry believes in Murati's vision.
But capital alone isn't enough to retain talent, especially in AI where options are abundant. OpenAI is clearly making a move to consolidate its research strength, pulling back researchers who'd ventured out to start something new. The pattern echoes what happened with John Schulman, who left OpenAI for Anthropic in 2024 before joining Thinking Machines as Chief Scientist.
The departures raise bigger questions about the sustainability of new AI startups founded by former executives of established labs. When the parent company wants talent back, and especially when that talent originally came from there, the gravity well is powerful. OpenAI has the resources, the compute power, and the existing infrastructure to make returning attractive. A startup, no matter how well-funded, has to prove it can compete on those fronts.
For Murati and the remaining team at Thinking Machines Lab, the challenge now is demonstrating that the company has a defensible mission and that losing key founders isn't a fatal wound. Promoting Soumith Chintala as the new CTO suggests the company sees internal talent it can rely on. But the back-to-back departures of founding team members will inevitably raise concerns from investors and potential partners about the company's stability.
The departure of Zoph, Metz, and Schoenholz represents a significant reset for Thinking Machines Lab at a critical moment in its development. While the company had the resources and investor backing to succeed, it couldn't retain the team that built its credibility. For the broader AI ecosystem, it's a reminder that even billion-dollar valuations can't guarantee stability when founders face the siren call of returning to the labs where they made their mark. OpenAI consolidates talent and maintains its research depth, while Thinking Machines Lab must prove it can still execute its vision with a reorganized leadership team.
