Andrew Tulloch, Meta’s star AI recruit, is leaving after less than a year

Andrew Tulloch is leaving Meta less than a year after he joined. His reported pay offer from the company had made him a symbol of the AI talent war. Reed Albergotti first reported the departure for Semafor on Wednesday, citing a person briefed on the matter. The Wall Street Journal followed on Thursday and reported that he is joining Anthropic.

Neither company has confirmed the move, and TNW has not independently verified it. The Journal’s Erin Woo and Meghan Bobrowsky cited people familiar with the situation. On Thursday, Tulloch’s own personal website still listed him at Meta’s TBD Lab, “working on superintelligence.”

He waited for Muse

According to Semafor, Tulloch held off leaving until Meta had shipped Muse, the personal AI agent it launched on Tuesday. He worked in TBD Lab, the small team at the centre of Mark Zuckerberg’s AI strategy. Scale AI founder Alexandr Wang leads it. No outlet has reported why he is going.

Tulloch knows Meta well. He spent 11 years there before leaving in 2023 for OpenAI. His website says he helped train GPT-4o, GPT-4.5 and o3 there. He then co-founded Thinking Machines Lab with Mira Murati, OpenAI’s former chief technology officer.

The offer Meta disputes

His return was a prize for Zuckerberg, who reportedly recruited him in person. Meta had first tried to buy Thinking Machines outright, an approach Murati rejected, Calcalist reported. Last year the Journal reported that Meta’s package could have reached $1.5bn over six years. That assumed bonuses and stock paid out in full.

Meta has always disputed that number. Its spokesperson Andy Stone called it “inaccurate and ridiculous,” as Gizmodo recalled. Tulloch turned the offer down at first and joined months later. Nobody has disclosed the sum he finally accepted.

Why Anthropic matters

If the Journal is right, the timing is awkward for Meta. It has spent heavily to rebuild its AI bench. That included a talent raid on Thinking Machines that brought over five of its founding members this spring. Now it loses the best-known name from that effort, a day after its biggest consumer AI launch. That raises the question of how long money alone can hold a lab together.

For Anthropic, it would be a marquee hire in the same week it lost a researcher in public. Jacob Coxon quit on Tuesday, warning that AI labs are gambling with our lives. TNW reported in August that Anthropic already pays the biggest salaries in the industry. It can bid for people few others can afford.

The churn is wider than two labs. Google chief scientist Jeff Dean left last month with three other senior researchers to start a company, Quartz reported. Only a small group of people have trained frontier models. Each move between labs carries that experience to a rival.

Neither Meta nor Anthropic has commented publicly on the move. Until one of them does, the only public record of where Tulloch works is his own website. It still says Meta.