Barret Zoph has joined Google as vice president of research, where he will work on reinforcement learning and post-training for Gemini. The move ends a five-month second stint at OpenAI that began in January 2026 and closes in June, and marks Zoph's third employer in under two years. Google confirmed the hire to the Wall Street Journal, framing it as a return — Zoph worked at Google earlier in his career before leaving for OpenAI.
Zoph's path reads like a case study in AI-executive churn. He spent two years at OpenAI before leaving in October 2024 to co-found Thinking Machines with former OpenAI CTO Mira Murati, who had exited the lab the month prior. Thinking Machines launched with one of the most-watched founding teams in the industry, positioning itself as a research-heavy competitor to OpenAI and Anthropic.
That plan did not hold. In January 2026, Zoph and fellow Thinking Machines co-founder Luke Metz departed the startup — sharply, and publicly — to return to OpenAI. Zoph was placed at the head of AI enterprise sales, an unusual assignment for a researcher whose reputation was built on model training rather than go-to-market work. He left again in June.
Key facts
- 01Barret Zoph joined Google as VP of research, assigned to reinforcement learning and post-training work on Gemini.
- 02Zoph spent only 5 months back at OpenAI after rejoining in January 2026, exiting in June to head to Google.
- 03He originally left OpenAI in October 2024 after 2 years to co-found Thinking Machines with Mira Murati.
- 04OpenAI has lost a string of senior staff over the past 8 months, including its COO and a top data center executive.
Google's public framing of the hire centers on capability, not politics. A Google spokesperson told the Wall Street Journal the company looks forward to Zoph bringing his reinforcement learning and post-training expertise to Gemini. Post-training — the RLHF and fine-tuning work that shapes how a base model actually behaves for users — is the layer where Google has visibly closed ground on OpenAI over the past year, and adding a senior researcher with hands-on frontier experience at two competing labs is a concrete lift for that team.
“We look forward to Barret returning to Google and bringing his RL and post-training expertise to Gemini.”— Google spokesperson, Google
The hire also lands in the middle of a persistent talent drain at OpenAI. The company has lost a string of high-level executives over the past eight months, including its chief operating officer and one of its top data center executives. OpenAI is preparing for an IPO and remains one of the most commercially dominant companies in AI, which makes the pattern of senior exits harder to explain as simple career progression.
AI Chat Daily covered part of that pattern last week in a piece on Greg Brockman consolidating control at OpenAI as executive exits mounted. Zoph's departure fits the same shape: senior technical and operational leaders leaving despite the company's commercial position, and landing at direct competitors rather than retiring or starting something new.
For Google, the recruiting win compounds. Gemini has been closing benchmark gaps with GPT-class models through 2026, and DeepMind has been public about scaling its post-training investment. Zoph's résumé — early work on neural architecture search at Google, then two years inside OpenAI, then a co-founder seat at Thinking Machines — gives him a rare view of how three different research cultures approach the same problems.
For Thinking Machines, the story is less flattering. Losing two co-founders back to OpenAI in January, only to watch one of them leave OpenAI again within five months and go to Google, raises questions about what the startup's technical direction looks like without them. The company has not publicly commented on Zoph's latest move.
There are limits to what one hire signals. Zoph is a strong researcher, but Gemini's trajectory is set by hundreds of engineers and by the compute and data pipelines behind them, not by any single VP. And executive movement at this level in AI has become common enough that a single transition rarely reorders the competitive picture on its own. OpenAI has not commented on Zoph's exit or on the broader pattern of departures.
The read for the AI market is that the frontier lab labor pool is smaller than the headcount charts suggest, and the same two dozen names keep rotating between OpenAI, Google, Anthropic, and a shrinking bench of well-funded startups. Google is currently winning that rotation on the post-training side, which is precisely the layer where Gemini needs to keep pressing to convert benchmark parity into product parity. Every senior researcher Google pulls out of OpenAI's orbit is one fewer person available to ship the next OpenAI model on schedule.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
