Mira Murati

there's a particular kind of credibility that only comes from having built the thing you're now arguing against — and right now, mira murati has it in full.

she grew up in Albania and later Canada, studied aerospace engineering at Dartmouth, and then made the kind of career pivots that in retrospect look inevitable: a stint at Tesla working on product, then a move to OpenAI back when it was still a research lab with an uncertain future. she rose to CTO, became the person who shepherded ChatGPT from internal experiment to the thing that rewired how the world thinks about software, and then — quietly, in September 2024 — she was gone. no scandal, no drama anyone could fully decode. just a departure that left a lot of people watching to see what she'd do with the runway.

what she's done is launch Thinking Machines Lab, and this week it became a lot more than a name.

murati just shipped Inkling — a 975-billion-parameter open-weight model, which is a genuinely significant number — and the argument embedded in that release is more interesting than the specs. the model is designed to be customizable, deployable by companies on their own infrastructure, adaptable to specific use cases rather than handed down as a finished product from on high. it's reported to match flagship performance at roughly a third of the token cost, which is the kind of claim that either holds up under pressure or doesn't, but either way it's a direct challenge to the closed, proprietary approach that OpenAI — the place she spent years — has committed itself to.

the timing lands in a specific cultural moment in AI. there's a genuine, unsettled argument happening right now about whether the closed-garden model or the open-weight model wins long-term — not just commercially but philosophically, in terms of who gets to build, who gets to customize, who retains leverage. Alex Karp and Satya Nadella have been out this week warning companies that open-sourcing AI is essentially giving away competitive IP. murati's answer, implicitly, is that the leverage isn't in the weights — it's in what you do with them. that's a bet, not a certainty, but it's a coherent one, made by someone who watched the closed approach operate from the inside at its highest levels.

Fast Company noted that Thinking Machines is coming for Anthropic's intellectual vibe — which is a pointed way of saying murati isn't just competing on benchmarks. there's a positioning play here, something about rigor and trust and who the serious researchers want to work with, that maps less onto the OpenAI playbook and more onto the careful, considered image Anthropic has built. whether Thinking Machines can actually occupy that space is an open question. but the fact that the comparison is being made at all suggests the debut registered.

what makes murati interesting to watch isn't the horsepower of Inkling or even the open-weight bet on its own. it's that she's one of a very small number of people who shaped the dominant paradigm in AI at close range, left it, and is now building an explicit counter-argument with her own capital and her own name on the door. that's a different kind of stake than most founders carry. the credibility and the vulnerability are the same thing.