Lilian Weng, a co-founder of Thinking Machines Lab, announced her departure this week, citing health reasons and a desire for clearer boundaries than the founder role allowed. The lab was created in 2024 by Mira Murati after she left OpenAI. Days after Weng's announcement she reappeared at OpenAI, now leading its recursive self-improvement (RSI) team — OpenAI's high-priority effort to develop models that can train their own successors.
Return timeline and numbers
Thinking Machines was founded by six people; four of those co-founders have left the lab, and three of them have returned directly to OpenAI. The startup raised a $2 billion seed round at a $12 billion valuation in 2024. During roughly eighteen months of operation, the lab did not release a widely recognized flagship model.
Why this matters
Thinking Machines was initially framed as a breakaway: after Mira Murati's exit from OpenAI, the new lab promised to pursue frontier AI in a setting that its founders portrayed as freer. The subsequent departures and the return of several founders to OpenAI indicate that much of the company's value was concentrated in its people, and that those people ultimately moved back toward the mothership.
Observers note that the episode looks less like a hostile poaching or dramatic collapse and more like a gradual reabsorption: the celebrated exodus has effectively become a paid sabbatical or an escape hatch quietly reattached to OpenAI. According to accounts, Sam Altman did not need to actively recruit the founders back; the movement unfolded without a direct campaign to poach them.
Conclusion
The Thinking Machines case highlights how founder-centered research ventures can be fragile even with large early funding. Despite a $2 billion seed raise and a high valuation, the absence of a flagship product and the departures of key personnel raise questions about the sustainability of startups built primarily around high-profile individuals.
