Picture this: you’re a senior researcher at Thinking Machines Lab. Your equity is tied to a company valued at $12 billion, founded by Mira Murati — the former CTO of OpenAI. You’ve got access to Nvidia’s latest GB300 chips through a freshly signed multibillion-dollar cloud deal with Google. Then your phone buzzes. It’s a recruiter from Meta.
Some people take that call. Some don’t. And that tension — who stays, who goes, and what it means for the bots we’re all building — is exactly what I want to dig into here.
A Two-Way Street With Unequal Traffic
The narrative that’s been floating around is simple: Meta is raiding Thinking Machines Lab. And yes, that’s happening. Mark Jen and Yinghai Lu are among at least three TML employees who have recently moved to Meta. When a company the size of Meta comes knocking with its resources and reach, some researchers are going to answer.
But the story doesn’t stop there. The talent flow runs both ways. Meta has also been losing people to TML, drawn by the startup’s momentum, its equity upside, and the kind of focused research environment that a company of Meta’s scale simply can’t replicate. Big tech has gravity. But so does a well-funded startup with a clear mission and a founder who helped shape one of the most influential AI labs in the world.
As someone who spends most of my time building bots and thinking about the infrastructure behind them, I find this dynamic genuinely interesting — not just as industry gossip, but as a signal about where serious AI work is actually happening right now.
What the Google Deal Actually Means
TML’s multibillion-dollar cloud deal with Google is the part of this story that deserves more attention than it’s getting. Securing early access to Nvidia’s GB300 chips puts TML in a very small group of organizations with the compute to train and run frontier models at scale. That’s not a minor footnote — that’s the kind of infrastructure advantage that shapes what’s possible.
For bot builders, this matters. The models that power the bots we ship are only as good as the training runs behind them. When a lab has access to next-generation hardware before most of the industry, it can iterate faster, experiment more aggressively, and push capability ceilings that the rest of us are still bumping our heads against. TML having that access means the models coming out of that lab — and eventually trickling into APIs and tooling we use — could move faster than expected.
Why Talent Flows Tell You More Than Press Releases
I’ve learned to watch where researchers move more than I watch product announcements. Press releases are optimized for perception. Talent decisions are made by people with real information, real options, and real skin in the game.
When experienced researchers leave a $12 billion startup for Meta, it tells you something about what Meta is building and how seriously it’s investing in that work. When other researchers leave Meta for that same startup, it tells you something about TML’s culture, its trajectory, and the kind of problems it’s letting people work on.
Both things can be true simultaneously. Meta is a serious AI organization. TML is a serious AI organization. The fact that talent is moving in both directions isn’t a sign of chaos — it’s a sign that both places are doing work compelling enough to attract people who have choices.
What This Means If You’re Building Bots Today
Practically speaking, here’s how I’m thinking about this as someone in the trenches:
- TML’s compute access via the Google deal suggests their model development timeline could accelerate. Worth keeping an eye on what they release and when.
- Meta’s continued investment in pulling top talent means its open-weight model strategy — which has already given bot builders a lot to work with — isn’t slowing down.
- The competition between these organizations, and others like them, is ultimately good for people building on top of their work. More competition means faster progress and more options.
The talent shuffle between Meta and Thinking Machines Lab is more than an HR story. It’s a map of where ambition and resources are concentrating in AI right now. And for those of us building on top of what these labs produce, that map is worth reading carefully.
Neither side is losing in any permanent sense. But TML, with its Google-backed compute deal and its founder’s pedigree, looks like it’s building something that serious people want to be part of. That’s usually a good sign of what’s coming next.
🕒 Published: