Meta's $1.5B AI hire Andrew Tulloch exits after ~11 months, joining Anthropic
量子位 · wechat · 2026-09-11
- Andrew Tulloch is leaving Meta's TBD Lab and joining Anthropic's inference & performance team to optimize Claude Code training, per reports — just two days after Meta shipped its Muse personal AI agent.
- Tulloch spent 11 years at Meta (PyTorch, ResNet-50, FBGEMM), moved to OpenAI to help train GPT-4o/4.5/o3, then co-founded Thinking Machines Lab.
- Last year Zuckerberg and Alexandr Wang personally courted him with a package reportedly worth $1-1.5B; he first refused, then rejoined Meta in Oct 2025.
- His second stint lasted only 11 months, echoing Jiahui Yu, hired at a reported $700M salary, who quit eight days after MuseSpark 1.2 shipped.
- Axios's framing: even sky-high pay, titles and compute can't retain top researchers; the talent war may cool only when elite experience becomes a replicable organizational engineering capability.
Related event: Ex-Meta Engineer Andrew Tulloch Joins Anthropic(3 posts)→
More from Companies & People
- Anthropic IPO filing reveals 5-7% of revenue came from rule-breaking customers, at least $814M — henkvaness · 2026-10-27
- LlamaIndex hosts AI leaders for three days at F1 Austin — llama_index · 2026-10-27
- Anthropic to be tightly controlled public company after IPO, investor applauds — matt_slotnick · 2026-10-27
- Alpha School head denies students skip books: 30-minute daily reading, a book every two weeks — nateliason · 2026-10-27
- Michael Dell: Musk is "the most demanding customer there is" — Colossus servers built in ~6 weeks — elonmusk · 2026-10-27
- Anthropic's $80.7B H1 loss: two customers drive a fifth of its revenue — tekbog · 2026-10-27