AI Researcher Ditches $1.5B Meta Payday For Robot That Might Kill Us All

KEY POINTS

•After less than a year at Meta, Andrew Tulloch announced he is joining AI startup Anthropic next week to optimize their AI models’ training and response efficiency.
•Tulloch’s career journey includes 11 years at Meta, a stint at OpenAI where he helped train GPT-4o and GPT-4.5, and cofounding Thinking Machines Lab in 2025.
•Anthropic is prepping for an IPO amid competition and public fears about AI’s risks, a concern voiced by former employee Jacob Coxon who recently quit citing existential danger.

Andrew Tulloch, the AI industry's version of a hot potato, has bounced from Meta to OpenAI, then his own startup, Thought Machines Lab, only to land at Anthropic this September. After originally labelling a $1.5 billion Meta offer as 'nah', he took a brief 11-year round-trip to Meta before sprinting off again. At Anthropic, where IPO vibes mix with apocalyptic AI anxiety—sparked by ex-staffer Jacob Coxon's viral warning that their creations might obliterate humanity by 2030—Tulloch plans to pimp AI training and inference. Meanwhile, Meta quietly launched Muse, an AI agent to mind your tasks, just as their star researcher slips out the back door.