Thinking Machines Lab (TML) has secured a significant workforce boost by recruiting top talent from Meta, including key figures in AI and deep learning research. This trend highlights an intense talent war between the established tech giant and the rising AI startup. Weiyao Wang, who spent eight years at Meta building multimodal perception systems and contributing to the SAM3D open-world segmentation project, recently joined TML following his departure from Meta last week. He joins a wave of former Meta employees moving to the startup, which includes Kenneth Li, a Harvard PhD who briefly worked at Meta before making the switch this month. TML is aggressively hiring from Meta, with recent data suggesting it has recruited more researchers from Meta than from any other single employer. The leadership exodus from Meta to TML includes prominent industry figures. Soumith Chintala, who spent 11 years at Meta and co-created PyTorch, left the company in late 2025 to become TML's Chief Technology Officer earlier this year. Similarly, Piotr Dollár, an 11-year Meta veteran and co-author of the influential Segment Anything model, has joined TML's technical staff. Other recent arrivals from Meta include research scientist Andrea Madotto, who focuses on multimodal language models, and software engineer James Sun, who has nearly nine years of experience in large language model training. This recruitment drive occurs as TML expands its technical capabilities and infrastructure. On Tuesday, at Google Cloud Next, TML announced a multibillion-dollar cloud agreement with Google. This deal provides the startup with access to Nvidia's latest GB300 chips, placing TML in the same high-end infrastructure tier as competitors like Anthropic and Meta. While Meta was reportedly interested in acquiring TML earlier this year and has attempted to hire away its founders, the startup continues to poach Meta's talent in return. Beyond its former colleagues, TML has attracted experts from across the industry. The roster includes Neal Wu, a former founding member of Cognition and International Olympiad in Informatics medalist; Jeffrey Tao, who arrived from Waymo, Windsurf, and OpenAI; and Muhammad Maaz, a former research fellow at Anthropic. Additionally, Erik Wijmans joined from Apple, and Liliang Ren came from Microsoft's AI Superintelligence team. The company's headcount now stands at approximately 140 employees. Despite being a relatively new venture that has released only one product so far, TML is valued at $12 billion. While this valuation may have been unimaginable for a company at this stage in previous technology cycles, it offers significant financial upside compared to the record-breaking valuations of OpenAI and Anthropic. The allure of TML is further strengthened by Meta's well-known seven-figure compensation packages, making the move an attractive option for researchers weighing their career options. A spokesperson for TML declined to comment on the specific recruitment details when contacted.