Barret Zoph, co-founder of the artificial intelligence start-up Thinking Machines Lab, has joined Alphabet’s Google as vice president of research, the company announced this week. Zoph’s return to Google follows a series of high-profile moves in the AI industry amid ongoing shifts in leadership and strategy.
Zoph originally co-founded Thinking Machines Lab in early 2025 alongside Mira Murati, CEO of the start-up, and four other partners. The company sought to develop AI systems designed to collaborate with humans rather than function autonomously. However, in January 2026, Zoph abruptly left Thinking Machines along with fellow co-founder Luke Metz and another employee, Sam Schoenholz, to rejoin OpenAI. All three had previously worked at OpenAI prior to their stint at the start-up.
At OpenAI, Zoph served as chief technology officer and was responsible for advancing the company’s AI products for business applications, including leading efforts around Codex, an AI platform designed to generate computer code. He was also involved in post-training processes, a technique for refining AI models after initial development.
Relations between Zoph and Murati deteriorated during their time at Thinking Machines. Murati reportedly fired Zoph citing issues related to performance, conduct, and trust, including concerns stemming from an undisclosed relationship with another employee, leading to a reduction in his responsibilities. Zoph contested these claims, stating that his dismissal came only after he informed Murati of his plans to leave and that any cutbacks in his role were temporary.
In addition to Zoph’s departure from OpenAI, sources indicate that Metz has also left the company and is joining Meta, reflecting continued fluidity among AI researchers between major tech firms.
Zoph originally worked at Google between 2016 and 2022 before joining OpenAI. His return to the company coincides with Google’s recent restructuring of its AI division and efforts to bolster its capabilities in areas such as reinforcement learning and post-training optimization. Thomas Schoenfelder, a Google spokesperson, highlighted that Zoph’s expertise aligns with the company’s focus on its Gemini AI project.
Google has been striving to catch up with competitors like Anthropic and OpenAI in developing AI systems capable of generating code, which remains one of the most commercially promising applications of artificial intelligence. The company’s leadership has also shifted recently, with Demis Hassabis stepping down as CEO of Google DeepMind last month to assume a chairman and chief scientist role within Alphabet, Google’s parent company.
Zoph’s move reflects the ongoing competition for talent among leading AI firms as they race to build increasingly advanced and commercially viable AI technologies.
