In a stunning AI talent war development, Barret Zoph has returned to OpenAI just 58 minutes after his exit from Thinking Machines Lab was announced. The move, alongside two other former colleagues, follows a reported amicable split, with OpenAI stating they “do not share” concerns about misconduct. This represents a significant blow to Thinking Machines’ ambitious plans.

Barret Zoph’s Wednesday went from career disaster to triumphant return in under an hour. The Thinking Machines Lab co-founder was fired by CEO Mira Murati, only to be welcomed back at OpenAI—where he’d worked before—just 58 minutes later.Murati’s terse announcement that she had “parted ways” with Zoph barely hit social media before OpenAI‘s Fidji Simo posted her own update: Zoph was rejoining the ChatGPT-maker, along with fellow Thinking Machines co-founder Luke Metz and researcher Sam Schoenholz. Simo told staff the move had been weeks in the making, with Zoph telling Murati on Monday he was thinking of leaving, according to Information.

The messy truth behind a very public split

What really happened depends on who you ask. Sources told Wired that Thinking Machines fired Zoph for “unethical conduct,” including allegations he leaked confidential information to competitors. The startup’s leadership reportedly lost trust in him after an earlier incident last year that soured his relationship with Murati, UDN reported.But OpenAI isn’t buying it. In her internal memo, Simo dismissed concerns about Zoph’s ethics entirely. She framed the departure differently: Zoph wanted out, Murati fired him Wednesday, and OpenAI swooped in. The timing raises eyebrows—Thinking Machines allegedly discovered the information-sharing concerns around the same time they learned about his OpenAI return.

Thinking Machines loses its third co-founder in months

For Murati’s startup, this stings. Thinking Machines raised $2 billion at a $12 billion valuation less than a year ago, positioning itself as a serious OpenAI rival. Now it’s lost three of its four co-founders. Andrew Tulloch jumped to Meta in November, and with Zoph and Metz gone, only Murati remains from the original team.She’s handed the CTO role to Soumith Chintala, a respected AI veteran. But losing your technical co-founders—especially one who led OpenAI’s post-training teams and another who worked on ChatGPT and the o1 model—is a blow no press release can spin away. Bloomberg reports at least two more Thinking Machines staffers are OpenAI-bound.The whole saga feels distinctly familiar. Murati herself was central to Sam Altman’s chaotic 2023 ouster from OpenAI. Now she’s on the other side of the drama, watching her company hemorrhage talent to her former employer.