Meta’s talent pipeline is leaking, and Thinking Machines is happy to catch it

Weiyao Wang spent eight years at Meta. That’s his entire post-college career. He worked on multimodal perception systems and contributed to SAM3D, Meta’s open-world segmentation project. His last day was last week. Now he’s at Thinking Machines Lab.

TML is having a moment. The startup just closed a multibillion-dollar cloud deal with Google, which gets them access to Nvidia’s GB300 chips — making them one of the first startups to run on that hardware. That announcement came at Google Cloud Next this past Tuesday. It follows an earlier partnership with Nvidia and puts TML’s infrastructure in the same tier as Anthropic and Meta.

Funny thing about Meta: reportedly held talks to acquire Thinking Machines around this time last year. Instead, they’ve been picking off TML’s founders one by one. But the talent flow runs both ways.

Kenneth Li — Harvard PhD, spent ten months at Meta — joined TML this month. Business Insider reported last week that Meta has poached seven of TML’s founding members. But if you look at LinkedIn profiles, TML has been hiring more researchers from Meta than from any other single company.

The biggest catch is Soumith Chintala, TML’s CTO. He spent eleven years at Meta, co-founded PyTorch — the deep learning framework that basically runs the world’s AI research now. He left Meta in late 2025 and became CTO earlier this year. Piotr Dollár is also on board: another eleven-year Meta veteran, former research director, co-authored the Segment Anything model. Andrea Madotto came from Meta’s FAIR division in December, working on multimodal language models. James Sun, nearly nine years at Meta working on LLM pre- and post-training, made the jump too.

TML isn’t just a Meta reunion though. Neal Wu — three-time gold medalist at the International Olympiad in Informatics, founding member of Cognition — joined early this year. Jeffrey Tao came via Waymo, Windsurf, and OpenAI. Muhammad Maaz held a research fellowship at Anthropic. Erik Wijmans arrived from Apple. Liliang Ren spent two and a half years on Microsoft’s AI Superintelligence team pre-training OpenAI models for code before joining in March.

Headcount is around 140 now.

Meta’s pay packages are legendary — seven figures, no strings attached. But for researchers weighing options, the calculus might be simpler than it looks. Thinking Machines Lab is valued at $12 billion. That figure would’ve been unimaginable for a company at this stage in any previous tech cycle — they’ve released exactly one product so far. Compared to OpenAI and Anthropic’s record-breaking valuations though, there’s still a lot of financial upside.

A spokesperson for TML declined to comment. Figures.