A Revolving Door With a Google-Sized Prize on the Other Side

Think of it like a busy restaurant kitchen. Chefs leave for the new place down the street, then some come back, then a few more leave again. From the outside it looks chaotic. But if that new place just landed a catering contract worth billions, you start to wonder who’s really winning the night.

That’s roughly what’s happening between Meta and Thinking Machines Lab right now, and as someone who spends most of his time evaluating AI tools and the teams building them, I find the dynamic genuinely fascinating — not just as industry gossip, but as a signal about where serious AI development is actually heading.

Two-Way Traffic, One Clear Momentum Shift

Meta has been pulling talent from Thinking Machines Lab. Mark Jen and Yinghai Lu are among the names that have moved over to Meta’s side. That’s real. Meta is a giant with resources, stability, and the kind of compensation packages that are hard to turn down.

But TML has been pulling talent from Meta too. Researchers have been drawn to TML’s $12 billion valuation and the equity upside that comes with being early at a startup that’s clearly on the rise. When you’re reviewing AI toolkits for a living, you learn quickly that the teams behind the tools matter as much as the tools themselves. A product is only as good as the people iterating on it, and right now TML is attracting people who want a piece of something they can help shape from the ground up.

So yes, Meta is losing people. And yes, Meta is gaining people. But the story isn’t really about headcount.

The Google Deal Changes the Calculus

TML recently signed a multibillion-dollar cloud deal with Google. That deal gives TML access to Nvidia’s latest GB300 chips, putting it among the first organizations to work with that hardware at scale. That’s not a minor footnote — that’s infrastructure that most AI labs would trade a lot to get their hands on.

From a toolkit reviewer’s perspective, compute access is everything. The difference between a tool that feels sluggish and one that feels sharp often comes down to what’s running underneath it. When a startup secures this kind of hardware partnership early, it compresses the timeline between research and usable product. Things that would take months to test can get validated in weeks.

TML was founded by Mira Murati, who previously served as CTO at OpenAI. That pedigree matters, but pedigree alone doesn’t build products. The Google deal is what turns ambition into actual capability.

What This Means for the Tools Coming Out of Both Camps

I review AI toolkits. I care about what ships, not just what’s announced. And when I look at this talent churn through that lens, a few things stand out.

• Meta’s ability to attract TML talent suggests it’s still a destination for people who want scale and distribution. Tools built at Meta reach a lot of users fast.
• TML’s ability to attract Meta talent — and retain enough people to keep growing — suggests it’s building something that researchers find worth betting on personally.
• The Google cloud deal means TML now has the compute to back up whatever it’s promising those researchers.

For anyone building on top of AI infrastructure, or evaluating which platforms to trust for serious work, this is the kind of signal worth tracking. A startup that can sign a multibillion-dollar deal with Google while simultaneously competing with Meta for talent is not a startup that’s struggling to find its footing.

My Honest Take

I’ve seen a lot of AI startups come through this space with big valuations and thin substance. TML doesn’t look like that from where I’m standing. The talent flow is real in both directions, which means Meta respects what TML is building enough to recruit from it — and TML is compelling enough that people leave Meta to join it.

That’s a genuinely healthy sign. Not because churn is good, but because it means both organizations are being pushed. Competition for talent forces better products, better culture, and faster iteration. The people caught in the middle — the researchers making these moves — are ultimately the ones who will determine what tools the rest of us get to use.

Watch what TML ships next. With GB300 access and a team that’s been hardened by this kind of pressure, whatever comes out of that lab is going to be worth a serious look.

🕒 Published: April 25, 2026