What if the most aggressive hiring machine in AI is accidentally building its own competition? That’s the uncomfortable question sitting at the center of the Meta-Thinking Machines Lab story — and it’s one that most coverage has been too polite to ask directly.
Here at agnthq, we don’t do polite. So let’s get into it.
The Setup
Meta has been picking off talent from Thinking Machines Lab (TML) since early 2026. On the surface, that reads like a straightforward big-fish-eats-small-fish story. A well-resourced tech giant hooking engineers and founders away from a startup that can’t match the salary packages. Classic Silicon Valley gravity.
Except that’s not quite what’s happening here.
According to reporting from multiple outlets including Yahoo Finance and 4AiTimes, this is a two-way street. Meta reportedly held acquisition talks with Thinking Machines around late 2024 or early 2025. Those talks didn’t close a deal. Instead, what followed was a slow, ongoing extraction of key people — including founders — from TML’s ranks.
And yet analysts are predicting significant growth for Thinking Machines. Not despite the talent drain. Possibly because of it.
Why Losing People Can Be a Strange Kind of Win
This sounds counterintuitive, but stay with me. When a startup loses senior people to a giant like Meta, a few things happen that don’t show up in the obvious narrative.
- The startup gets forced to rebuild, often with hungrier, more focused talent.
- The people who leave carry institutional knowledge into a company that is now, in some ways, a satellite of the original vision.
- The startup’s profile rises. Suddenly everyone wants to know what TML is building that Meta wanted badly enough to poach its founders.
That last point is underrated. Attention in the AI space is a resource. When Meta circles your company — first with acquisition interest, then with a hiring vacuum — you become a name people track. Investors notice. Engineers notice. Customers notice.
The Chintala Signal
One data point worth paying attention to: Soumith Chintala, co-creator of PyTorch, was appointed CTO of Thinking Machines Lab in early 2026. That’s not a consolation hire. That’s a statement. PyTorch is the foundation that a significant chunk of modern AI development is built on. Bringing in someone with that kind of credibility and technical depth suggests TML isn’t limping along after losing people to Meta — it’s reconfiguring.
If you’re trying to read the tea leaves on where TML is headed, that appointment tells you more than any press release would.
Meta’s Side of This Is Less Flattering Than It Looks
From Meta’s perspective, this looks like smart talent acquisition. And maybe it is. But there’s a version of this story where Meta’s approach reveals a strategic gap rather than a strength.
The company reportedly tried to buy TML outright and didn’t. Then it started pulling people out one by one. That’s not a confident move from a company that knows exactly where it’s going. That reads more like a company that sees something it wants and can’t quite get its hands around it.
Meta has the resources to build almost anything. The fact that it was interested in acquiring TML — and then settled for poaching — suggests Thinking Machines had something genuinely difficult to replicate from scratch. That’s a meaningful signal about TML’s actual technical position, even if the org chart looks messier now than it did a year ago.
What This Means for the Broader AI Talent War
The Meta-TML dynamic is a clean example of something playing out across the AI space right now. Large incumbents are using their financial weight to pull talent from smaller labs, but the smaller labs are proving more resilient than expected. They adapt. They attract new people who want to work somewhere that isn’t a bureaucracy. They use the attention to raise capital.
For anyone watching the AI agent and tooling space — which is exactly what we do here — the lesson is that headcount is not the same as momentum. A startup that loses five senior engineers to Meta but gains a CTO with Chintala’s profile and a sharper focus is not a startup in decline.
Thinking Machines Lab looks, from the outside, like a company that got raided. Look closer and it looks more like a company that got stress-tested and is still standing. Analysts predicting growth aren’t being optimistic. They’re reading the actual signals.
Meta wanted TML badly enough to try buying it. That tells you everything about who’s really winning this one.
🕒 Published: