Barret Zoph’s Third Pivot: From Thinking Machines to OpenAI, Now Joins Google

Core Announcement and Key Facts

On August 27, 2026, TechCrunch confirmed Barret Zoph has joined Google as Vice President of Research. Zoph will return to Google—where he previously worked—and focus on reinforcement learning (RL) and post-training techniques to support Gemini’s development, according to a Google spokesperson quoted by the Wall Street Journal.

Critical timeline and role transitions:

Before October 2024: Spent two years at OpenAI, then departed;
October 2024: Co-founded Thinking Machines with Mira Murati, serving as co-founder and CTO;
January 2025: Dramatically Left Thinking Machines alongside co-founder Luke Metz; later confirmed he was fired, then briefly rejoined OpenAI;
June 2025: Departed OpenAI (held role for five months, leading AI enterprise sales);
August 2026: Confirmed appointment at Google.

Industry Volatility Reflected in a Single Career Path

Zoph’s trajectory illustrates the exceptionally high mobility of AI executives in today’s landscape. He has cycled through three of the most prominent AI institutions—OpenAI (twice), Google, and a startup he co-founded—with limited tenure at each stop.

A key contradictory detail emerges: Zoph’s combined tenure at OpenAI totals roughly 2.5 years (2022–2024 + 5 months in 2025), yet his co-founder role at Thinking Machines lasted only approximately three months before termination. Such a rapid collapse of a startup venture, especially with a high-profile co-founder like Murati, remains uncommon in the AI ecosystem.

Zoph and Mira Murati, who left OpenAI’s AI lab in September 2024, launched Thinking Machines in October 2024. Just three months later, both Zoph and co-founder Luke Metz departed simultaneously. TechCrunch confirmed Zoph was fired.

OpenAI’s Executive Churn as an Industry Signal

Zoph’s career pattern mirrors broader talent instability at OpenAI, despite its IPO preparations and market dominance. Over the past eight months, OpenAI has lost senior leadership across functions—including COO, and recently a top data center executive—raising external questions about governance coherence.

Reinforcement learning (RL) refers to a machine learning paradigm where an agent learns optimal actions by interacting with an environment and receiving feedback in the form of rewards. It is now a cornerstone approach for_aligning_ large models with human intent.

Zoph’s return to Google explicitly targets RL and post-training domains, aligning with Google’s long-term bets on AGI (Artificial General Intelligence) research.

Guidance for Readers

For researchers and engineers: Those focused on RL, post-training, or model alignment should monitor Zoph’s upcoming hires, publications, and project shifts, which may indicate expanded internal capacity in these areas at Google.

For founders and job seekers: OpenAI’s executive retention challenges are intensifying. Zoph’s two departures hint at limitations in career stability—even at top-tier AI companies—where governance structures remain fluid despite scale. Founders also face heightened founder-foundering risk: startups can fracture quickly among leadership splits.

Final Note

Zoph’s career chain is both a personal journey and a professional barometer. As the large-model race enters an era of scaling and safety alignment, executive mobility has become a leading signal of technical direction shifts and organizational health.

(Note: This article strictly reconstructs facts from TechCrunch’s reporting; no unverified figures or speculative claims are added.)