Mira Murati.

Mira Murati was born in Vlorë, Albania, in 1988 — three years before the country opened to the West. She left for boarding school in Canada at sixteen, then enrolled at Dartmouth, where she finished a mechanical engineering degree in 2012. Her early career was hardware: a product role at Goldman Sachs, then a stretch at Zodiac Aerospace, then four years at Tesla as a senior product manager on the Model X. She moved to Leap Motion as head of product and engineering, where she helped pivot the company toward VR and AR. In 2018, she joined OpenAI as VP of applied AI and would, six years later, be the CTO under whose name DALL-E, ChatGPT, GPT-4, and Sora shipped.

Murati's tenure at OpenAI ran through every major inflection of the modern AI era — and every major internal controversy. She was the public-facing technical voice during the November 2023 board crisis that briefly removed Sam Altman. She faced a wave of press attention during the 2024 dispute with Scarlett Johansson over a voice that sounded like the actress in GPT-4o's voice mode. The leadership exodus that followed — Ilya Sutskever, Jan Leike, Mira herself in September 2024 — was widely read as a referendum on the lab's safety culture. The challenge of running Thinking Machines is, in the abstract, easier than running OpenAI at its peak; in practice, it means rebuilding the institutional muscle of a frontier lab from a standing start, with no product yet, in a market where the foundation-model layer is rapidly commoditizing.

Thinking Machines Lab announced its founding in February 2025. By July 2025, press reports placed a seed round at roughly $2B at a $12B post-money valuation — among the largest seed rounds in venture history. Investors included Andreessen Horowitz, Accel, Nvidia, AMD, and several others. The founding team is widely covered as a who's-who of ex-OpenAI alumni — John Schulman, Barret Zoph, and a long roster of senior researchers. The lab's positioning has been deliberately reserved: a research focus on alignment, multimodal models, and human-AI collaboration, with the first major product release projected for 2026.

Murati's public commentary, while sparse, has converged on a small set of ideas. That a research lab is fundamentally a company about people, not models — the cap-table negotiation matters less than the founding cohort. That the institutional architecture of a frontier lab determines whether its research stays on the same orbital as its product, and that OpenAI's drift was structural rather than personal. That safety and capability are not opposites — but that you only get to argue that publicly if you have built the capability side, which she has.