Top AI Researcher Andrew Tulloch Leaves Meta for Anthropic
Left 50%
Center 0%
Right 50%
1 left · 0 center · 1 right
What happened
Andrew Tulloch, a prominent AI researcher, is leaving Meta after less than a year to join Anthropic, where a company spokesperson said he will work on the inference and performance team to optimize model training and how systems respond to users. Semafor first reported the departure. Tulloch had spent 11 years at Meta, left for OpenAI in 2023 and helped train its GPT-4o, GPT-4.5, and o3 models, co-founded Thinking Machines Lab in 2025, and returned to Meta in October to work in Superintelligence Labs under Alexandr Wang. While he was at Thinking Machines Lab, The Wall Street Journal reported that he turned down a Meta pay package that could have totaled $1.5 billion over six years; Meta spokesperson Andy Stone called that report "inaccurate and ridiculous." The exit came shortly after Meta launched Muse, an AI agent meant to autonomously complete tasks for users.
Sourcing check · Mostly on the record · 2 articles
1 of 2 articles cite a named source for the story. It started with a Semafor report citing unnamed sources; a named source appeared 4 hours later. Most articles credit Semafor rather than reporting it themselves.
What these terms mean and how they're counted
- On the record · names a source 1 of 2
- On the record means the article cites a source by name for the story's main claim: a person, or an institution speaking officially, such as a company statement or a named agency's spokesperson. The opposite is an unnamed source. The count is how many of the counted articles cite at least one such source; the names are everyone named across them, each once however the articles spell it. Here: Anthropic spokesperson
- Primary evidence · cites a document or data 0 of 2
- Primary evidence means the article cites a record you could check yourself: a court filing or ruling, a dataset or report, a recording or transcript, an official statement or press release, market data. A quote is not a document. The count is how many articles cite one. Here: None cited.
- Credits another outlet 2 of 2
- The article says the story comes from another outlet's reporting (“first reported by Reuters”). Those articles relay the story; they are not independent confirmation. Here: Semafor
- Unnamed sources 0 of 2
- The main claim is attributed to people who are not identified: “people familiar with the matter”, “officials said”. The count is how many articles do this.
- How it developed
- When the first report landed and whether it named anyone; when the first named source and the first document appeared. Here: First report Sep 10, 17:12 UTC by Breitbart, crediting Semafor, with no named source. First named source 4 hours later, from Business Insider.
- Not counted 0
- Opinion, analysis and video pieces, and articles too short to assess. Every count above is out of the counted articles. An article republished under a new headline is one article.
This is about who is cited, not whether the story is true. Each source below says what it names. How we count
Omitted — what each side leaves out
Unpacked
Right-leaning coverage never names Tulloch's next employer, so the exit reads as a walkout rather than a hire by a named rival. Left-leaning coverage reports he is joining Anthropic next week, confirmed by a company spokesperson, to optimize training and how models respond to users. That fact is the difference between Meta simply losing a star and a direct transfer of model-training experience to a competitor preparing to go public.
One side casts a job hop, the other a personal snub. Left-leaning coverage says he "ditches Meta." Right-leaning coverage labels him Zuckerberg's "AI Golden Child." Those two labels decide whether this news is career mobility or humiliation.
What made Tulloch leave Meta so soon after he arrived in October?
Bottom line
If you read Breitbart, Tulloch just walked out on Zuckerberg after a year, with no next job. Business Insider has Anthropic confirming he starts there next week.
The receipts — 6 facts one side's coverage lacks
- Tulloch is joining Anthropic's inference and performance team next week, confirmed by an Anthropic spokesperson
- Meta spokesperson Andy Stone called the $1.5 billion pay-package report inaccurate and ridiculous
- At OpenAI, Tulloch helped train GPT-4o, GPT-4.5, and o3
- Tulloch worked in Meta's TBD lab within Superintelligence Labs and reported to Alexandr Wang
- Tulloch had worked at Meta for 11 years before leaving for OpenAI in 2023
- Zuckerberg tried to acquire Thinking Machines Lab, then approached its staff directly
The Left View
Business Insider presents the move as part of ongoing "AI's talent wars," in which experienced researchers have "a great deal of leverage — and earning power." It stresses that Tulloch will join Anthropic's inference and performance team and argues the hire "may offer some calm in an intense moment" as Anthropic prepares for an initial public offering, competes with rivals for business, and reckons with "mounting concerns about the adverse effects of AI." That framing is linked to researcher Jacob Coxon's resignation from Anthropic and his statement that "the people building AI earnestly believe that it could kill us all by the end of the decade."
The Right View
Breitbart frames the story as Mark Zuckerberg's "AI Golden Child" walking away after a reported mega-package lure, and notes that Tulloch "has not publicly announced why he left Meta or where he's headed next." It recounts Zuckerberg's unsuccessful attempt to acquire Thinking Machines Lab and his subsequent direct recruiting of its staff, and treats the short Meta tenure as part of industry-wide turmoil: "Zuckerberg isn't the only AI titan suffering a loss of talent," with OpenAI described as a "revolving door" after safety and alignment leaders left. The outlet concludes that AI giants face "high turnover as the industry engages in a multibillion-dollar game of chess."
Our Take (balanced)
The left-leaning account's strongest argument is that scarce AI talent now dictates terms among well-funded labs: Business Insider's confirmation of Tulloch's Anthropic role, plus the company's IPO timing and Coxon's claim that builders "earnestly believe" AI "could kill us all by the end of the decade," supports reading the hire as a competitive win and a bid for stability. The right-leaning account's strongest argument is the mismatch between Meta's aggressive, costly courtship and a sub-year stay, which Breitbart uses, alongside OpenAI safety-staff exits, as evidence that frontier labs are unstable despite vast spending. The central unresolved tension is whether this is primarily a story of researcher leverage in an open market or of failed high-stakes poaching and institutional churn inside the leading AI companies.
2 sources
More stories
The week's bottom lines, in your inbox
One email a week: the five stories that mattered and what they actually mean. Free.