Thinking Machines co-founder Zoph returns to Google as VP of research

Share
Thinking Machines co-founder Zoph returns to Google as VP of research

One of the most-watched names in frontier AI is going home. Barret Zoph, a co-founder of Thinking Machines Lab and a veteran OpenAI researcher, is returning to Google as vice president of research, the Wall Street Journal reports.

Thinking Machines Lab co-founder Barret Zoph is returning to Google as a vice president of research, the latest swing in the talent war behind the frontier model race. Zoph left OpenAI in June and stepped away from Thinking Machines Lab in January, according to the Journal. No specifics on his mandate or which Google team he lands in have been confirmed, but a VP-of-research title in Mountain View points squarely at Google DeepMind's emerging-model work rather than the search side of the business.

Zoph matters less for any single model than for what his movements say about how the industry's brains are reshuffling. He was part of the founding group of Thinking Machines Lab, the seven-person outfit that former OpenAI CTO Mira Murati launched in 2025 and that later pulled in Ilya Sutskever, becoming one of the most prominent challengers to the established labs. Losing a co-founder to a rival — and to Google specifically — is a real dent in that narrative, even after the lab promoted Soumith Chintala to CTO to steady the bench. For Google, the hire caps a period in which it has aggressively re-acquired and retained the researchers who left to chase the new generation of labs.

The return also resets a relationship that goes back further than most people remember. Zoph was at Google Brain before his OpenAI years, so "return" is literal, not just a headline verb. In practical terms, Google gains a researcher with deep experience in the post-training and RLHF-era techniques that now sit at the center of every frontier lab's roadmap — precisely the expertise DeepMind under CEO Kavukcuoglu says it is hoarding as the Gemini line pushes into agentic territory.

The take: talent is the cheapest real moat in AI, and Google just bought a piece of it back. None of this is publicly confirmed by Zoph, Google, or Thinking Machines Lab, so treat the specifics as widely reported but unconfirmed. Whether it's a homecoming or a poach depends on who's telling the story — but either way, the frontier churn isn't cooling.

Google has been on both ends of these swings lately — does re-acquiring founders beat building new talent? Tell us in the comments.

Read more

Byteification turns Qwen3 and Llama 3 into byte-level models

Byteification turns Qwen3 and Llama 3 into byte-level models

The tokenizer is the last piece of the language-model pipeline that was never learned end to end — and a Nature paper out this week shows it can be swapped out for less than one percent of a pretraining budget. Researchers have retrofitted four open subword language models into true byte-level models using less than 1% of a typical pretraining budget, and the result — published 7 October in Nature — closes the performance gap that has kept byte-level models a niche curiosity for nearly a decad

OpenAI's dots go mobile and can start your Codex work

OpenAI's dots go mobile and can start your Codex work

One product update and two benchmark papers this hour, all circling the same question: how much of the work can you actually hand over, and what happens to the part you didn't? OpenAI's dots — the always-on agents that get their own cloud computer — can now be created and configured entirely from the ChatGPT app on iOS and Android, and a dot can start Codex work or pick up an existing Codex thread on your behalf. The October 9 release, billed as the first of a fresh run of updates for dots, mo

Coding agents as a hospital: Cockroach's five-month results

Coding agents as a hospital: Cockroach's five-month results

The agentic-coding debate keeps being framed as speed versus trust. Cockroach Labs just published the most detailed trust-first ledger yet — five months of numbers, including the parts that didn't work. Cockroach Labs ran its codebase like a teaching hospital for five months and says the pipeline merged well over a million lines of code with seven total reverts. In the setup the company calls MOLT Sinai, GitHub issues are patients, merging is discharge, and the humans in charge are Chiefs of M

Today in AI — October 10, 2026

Today in AI — October 10, 2026

A day where the boring machinery took center stage: proof checkers, order books, warehouses and the humble text message — plus fresh arguments about who gets to declare any of it safe. Models & Research * Mathematicians get a reliability primer for the Lean Theorem Prover. A guest post by Thomas Hales on Terence Tao's blog walks through what Lean actually guarantees — and devotes itself to the "Summer of Soundness Bugs," the string of kernel bugs found this summer that let false proofs thr