I’m sure you’ve heard the breaking news: three senior figures left the Thinking Machines Lab, with some drama attached. That happens. Putting aside the obvious talent war that’s going on, leadership transitions, disagreements about direction, and public exits are not rare in frontier labs.
What surprised me wasn’t the departures themselves, but the reaction.
Scrolling through Twitter, I saw a lot of giddiness: people declaring the lab “finished,” predicting collapse, or arguing that its leadership lacks the “right” background to run a research organization. The whole thing was treated more like sport!
Now, this sits oddly with something you hear constantly from AI leaders: the best outcome for AI is many winners. Not one dominant lab. Not a single monopoly on progress. But a broad, competitive ecosystem where multiple entities push each other forward.
You can find versions of this idea all over the place:
- Jensen Huang has coined the term “Sovereign AI.” His argument is not just that companies should compete, but that countries must compete and possess independent capability.
- Anthropic’s “race to the top” argues that safety is best achieved by competition that incentivizes responsibility.
- Andrej Karpathy has been a vocal proponent of the “ecosystem” view.
- This is Altman on Fridman’s podcast:
“I think there’s going to be many AGIs in the world. So we don’t have to out-compete everyone. We’re going to contribute one. Other people are going to contribute some. I think multiple AGIs in the world with some differences in how they’re built and what they do and what they’re focused on… I think that’s good.”
- And LeCun on Fridman’s podcast:
“I see the danger of this concentration of power through proprietary AI systems as a much bigger danger than everything else.”
Here’s the contradiction: if we believe in a plural future for AI, why are we so eager to cheer when a lab stumbles?
A “race to the top” doesn’t mean much if there’s no one left to race against. Competition only works if we actually want competitors to survive, adapt, and occasionally surprise us.
That’s why I’m rooting for Thinking Machines. They’ve published some of the most thoughtful and well-written blog posts I’ve read, and Tinker helped me experiment with larger models (without emptying my bank account). I hope they keep building!
