Meta’s loss is Thinking Machines gain - BERITAJA

Weiyao Wang spent 8 years astatine Meta — his first occupation retired of assemblage — helping build multimodal cognition systems and contributing to open-world segmentation projects, including SAM3D. His last time astatine Meta was past week, and he has since joined Thinking Machines Lab.

His move to Thinking Machines Lab (TML) comes arsenic the AI startup expands connected aggregate fronts. TML conscionable signed a multibillion-dollar unreality deal pinch Google, giving it entree to Nvidia’s latest GB300 chips and making it 1 of the first startups to tally connected the hardware.

The agreement, announced this past Tuesday astatine Google Cloud Next, follows an earlier business pinch Nvidia, and puts TML successful the aforesaid infrastructure tier arsenic Anthropic and Meta. (Meta reportedly held talks to get Thinking Machines about this clip last year and has much precocious been picking disconnected TML’s founders 1 by one.)

The talent image remains fluid. Wang and Kenneth Li — a Harvard PhD who spent 10 months astatine Meta earlier joining TML this period — are the latest examples of a talent drawback that runs successful some directions. Business Insider reported past week that Meta has poached seven of TML’s founding members. A reappraisal of caller hires shows Thinking Machines is raiding Meta correct back. It appears, based connected a reappraisal of LinkedIn profiles, that TML has been hiring much researchers from Meta than from immoderate different azygous employer.

The about salient is Soumith Chintala, TML’s CTO, who spent 11 years astatine Meta and co-founded PyTorch, the open-source heavy learning model that now underpins about of the world’s AI research. He near Meta successful precocious 2025 and was appointed CTO earlier this year. Piotr Dollár, different 11-year Meta seasoned who served arsenic investigation head and co-authored the influential Segment Anything model, is now connected TML’s method staff. Andrea Madotto, a investigation intelligence successful Meta’s FAIR section focused connected multimodal connection models, joined TML successful December. James Sun, a package technologist pinch about 9 years astatine Meta moving connected LLM pre- and post-training, besides made the jump.

TML has drawn talent from beyond Meta, too. Neal Wu — a three-time golden medalist astatine the International Olympiad successful Informatics and a founding personnel of the buzzy coding startup Cognition — joined early this year. Jeffrey Tao came via Waymo, Windsurf, and OpenAI. Muhammad Maaz antecedently held a investigation fellowship astatine Anthropic. Erik Wijmans arrived from Apple. Liliang Ren spent 2 and a half years connected Microsoft’s AI Superintelligence squad pre-training OpenAI models for codification earlier joining successful March.

The startup’s headcount now stands astatine about 140.

Meta’s salary packages — 7 figures, nary strings attached — are good known by now. For researchers weighing their different options, the calculus whitethorn beryllium arsenic elemental arsenic this: Thinking Machines Lab is correct now weighted astatine $12 billion. Though that fig would’ve been unimaginable for a institution astatine this shape successful immoderate erstwhile tech rhythm (it has released conscionable one product truthful far), compared pinch the record-breaking valuations of OpenAI and Anthropic, there’s still a batch of financial upside.

Reached Friday morning, a spokesperson for TML declined to remark for this story.