Inside the turmoil at Thinking Machines; sources say Meta discussed buying TML, and CTO Barret Zoph had been in talks since October 2025 about an OpenAI return
Defections, secret conversations, deal talks that fizzled and a battle for control: The turmoil at Thinking Machines Lab …
Context & Ripple Effects
Thinking Machines Lab's internal strain had already surfaced in reports of allegations involving confidential company information and additional employees expected to join OpenAI. The reported Meta discussions place that personnel instability alongside a potential change in ownership or strategic control.
For Meta, the episode fits earlier reporting that its AI leadership had considered using rival models rather than relying solely on Llama. A deal that did not materialize nevertheless indicates that an independent frontier lab was at least part of Meta's strategic options.
First-order effects
- Thinking Machines Lab must manage continuity and credibility while its CTO's reported OpenAI discussions and wider defections put leadership stability under scrutiny.
- Meta does not gain control of TML from the reported talks, leaving both companies to pursue their AI strategies independently for now.
Second-order effects
- OpenAI gains leverage in recruiting from a lab already facing departures, while TML's remaining leadership may need to spend more effort on retention and governance than on external positioning.
- Meta's interest can raise the strategic value of independent AI-lab talent and research, even without an acquisition, increasing pressure on rivals to compete for both.
Third-order effects
- If repeated, such episodes would further concentrate frontier-AI capabilities in a small group of cash-rich platforms and established labs as independent teams become targets for hiring or acquisition.
- The pattern also makes internal governance a competitive variable for AI labs: talent retention and control of sensitive information can shape a lab's bargaining power as much as technical progress.
The trend: Frontier AI is moving toward tighter competition for scarce research teams, with platform companies using hiring and potential acquisitions alongside in-house model development.