What Mira Murati left OpenAI to build

On the day in September 2024 when Mira Murati announced she was leaving OpenAI, the press cycle wrote two stories simultaneously. One was about the exit itself — the chief technology officer of the most consequential AI lab in the world walking away at the peak of its public arc, joining a months-long roster of senior departures that included Ilya Sutskever and Jan Leike. The other was about what came next. By early February 2025, Murati had announced Thinking Machines Lab. By July, press reports placed the seed round at roughly two billion dollars on a twelve-billion-dollar post-money valuation. There was no product. There was no public paper. There was a website, a founding team, and a positioning statement.

The round was reported as among the largest seed rounds in venture history. The cohort that joined her — John Schulman, Barret Zoph, and a long roster of senior OpenAI alumni — was the kind of hire list that startups normally spend a decade assembling. Whatever Thinking Machines does next, the founding decision has already redrawn the talent map of frontier AI research.

The years that produced her

Murati was born in Vlorë, Albania, in 1988. She moved to Canada for boarding school at sixteen, then to Dartmouth, where she finished a mechanical engineering degree in 2012. Her first job out of school was a product role at Goldman Sachs; her second was Zodiac Aerospace. The path was hardware. The Tesla role came in 2014 — a senior product manager position on the Model X. She did about two years there before moving to Leap Motion as head of product and engineering, where she helped reposition the company from gesture controls toward VR and AR.

She joined OpenAI in 2018 as VP of applied AI. By the time she was elevated to CTO in 2022, the lab was already on the inflection that would, a year later, make ChatGPT the fastest-growing consumer product in the history of software. Over the four years she ran engineering and product, the lab shipped DALL-E (2021), ChatGPT (November 2022), GPT-4 (March 2023), GPT-4o with voice and vision (May 2024), and Sora (announced February 2024). She was the public-facing technical voice for most of those launches.

The November

The crisis that defined her tenure was the November 2023 board incident. Sam Altman was removed from his post on a Friday evening. Murati was named interim CEO on the Friday night. By Sunday, most of the lab's senior staff had signed a letter demanding Altman's return. By Tuesday morning, the board had reversed itself and Altman was back. Murati spent the four-day window as the most senior continuity person inside the lab; her handling of it was widely credited with preventing a more chaotic outcome.

What did not happen, in the months that followed, was a resolution of the underlying issues. The drift between the lab's research mission and its product surface remained. The safety team's relationship to the deployment team remained complicated. The pace of senior departures — Ilya Sutskever in May 2024, Jan Leike days later, Mira herself in September — read like a slow-motion exodus.

The founding move

Thinking Machines was announced on February 18, 2025. The launch post said almost nothing about the product. It said a great deal about people — that the lab was being staffed almost entirely from the senior research benches of OpenAI, Anthropic, and Google DeepMind, and that the focus would be on three areas: alignment, multimodal models, and human-AI collaboration. A research stance, not a product roadmap.

The seed round followed in July. The reported size — $2B at a $12B post — drew comparisons to Inflection AI's 2023 fundraise as the largest seed in the history of the category. Investors named in the round included Andreessen Horowitz, Accel, Nvidia, and AMD. The structure of the round was reported to include unusual governance terms protecting Murati's control of the lab, in a direct response to the OpenAI board structure that had backfired in 2023.

“

A research lab is a company about people, not about models. The cap table matters less than the founding cohort.

What the bet is

Strip away the press coverage and the founding bet is narrow. Murati believes that the institutional architecture of a frontier AI lab — the board, the charter, the incentive structure, the talent density — is what determines whether research stays close to product, and that OpenAI's drift between the two was structural rather than personal. Thinking Machines is an attempt to redesign that architecture from scratch, with the talent that proved it could ship at the previous lab.

The market backdrop is that the foundation-model layer is rapidly commoditizing. Anthropic, Google DeepMind, Meta's open-weight teams, and the Chinese frontier labs are converging on similar capabilities at similar costs. The economic value is migrating up the stack to applied surfaces — Perplexity, Cursor, Notion, ChatGPT itself — and to vertical applications. A new lab entering at the model layer has to either commit to enormous capex (matching the frontier) or find a research focus that bends the cost curve.

Thinking Machines' public stance is closer to the latter. The areas Murati has called out — alignment, multimodal, human-AI collaboration — are not the same as racing GPT-5. They imply a research culture that wants to ship narrower, better-aligned models rather than a single monolithic frontier capability. Whether that turns into a product the market will pay for, and whether that product can compete with the OpenAIs and Anthropics that are spending tens of billions a year, is the question the next eighteen months will answer.

What founders are watching

The Murati story has become a kind of organizational case study among other technical founders. Three lessons are usually pulled out of it.

The first is that a research lab is a company about people. Murati's founding was effectively a hiring exercise dressed up as a fundraise — the $2B was for the cohort, not the GPUs. The cap table matters less than the team you start with. Founders raising in adjacent categories now ask, as one VC put it, the Murati question: who's coming with you?

The second is that institutional architecture compounds. OpenAI's structural mismatch between its 501(c)(3) nonprofit board and its for-profit subsidiary was the proximate cause of the November 2023 crisis. Murati's reported governance design at Thinking Machines is meant to avoid that mismatch — clearer chain of decision-making, fewer veto points, the founder protected from the kind of board action that almost ended Altman's run. It's a quiet structural lesson that most consumer-AI founders are now studying.

The third is that safety and capability are not opposites. Murati shipped DALL-E, ChatGPT, GPT-4, and Sora before she ever publicly argued the safety side of the trade. That sequence matters. Founders who try to argue the safety side without first having built the capability side find it hard to be taken seriously. Founders who build first, then argue, get a different audience.

What's next

Thinking Machines has said publicly that its first major product release is targeted for 2026. The lab is hiring aggressively — both in San Francisco and remotely — and has been clear that its early years will be research-led rather than launch-led. The funding gives Murati a multi-year runway to do that without product pressure.

Whether the company becomes a generational research lab — an Anthropic-scale outcome, or larger — is not yet knowable. The founding team makes it plausible. The capital makes it possible. The market backdrop makes it harder than it has been for any new frontier lab in a decade. The next product release will be the first datapoint that settles which of those is the dominant force.

Quick takeaways

• Murati was the CTO under whose name DALL-E, ChatGPT, GPT-4, and Sora all shipped — the most consequential product run of any AI executive in the field.

• Thinking Machines' founding was a hiring exercise dressed up as a fundraise; the $2B was for the cohort, not the GPUs.

• The governance design of the lab is a direct response to OpenAI's November 2023 board crisis.

• The research focus — alignment, multimodal, human-AI collaboration — implies a narrower bet than racing GPT-5.

• The first major product release is targeted for 2026; the next 18 months will settle whether the bet works.

FAQ

Who is Mira Murati?

Founder and CEO of Thinking Machines Lab. Previously chief technology officer at OpenAI from 2022, where she led the launches of DALL-E, ChatGPT, GPT-4, and Sora. Born in Vlorë, Albania; mechanical engineering at Dartmouth.

Why did she leave OpenAI?

She announced her departure in September 2024 after nearly six years at the lab, joining a months-long roster of senior departures that included Ilya Sutskever and Jan Leike. Public statements emphasized a personal next chapter; press coverage read the exit as part of a broader drift between the lab's research mission and product surface.

What is Thinking Machines Lab?

An AI research lab announced in February 2025. The reported focus is alignment, multimodal models, and human-AI collaboration. The founding team is staffed largely from senior research benches at OpenAI, Anthropic, and Google DeepMind.

How much did Thinking Machines raise?

Press reports place the seed round at approximately $2B at a $12B post-money valuation, among the largest seed rounds in venture history. Investors included Andreessen Horowitz, Accel, Nvidia, and AMD. Exact figures may move with subsequent rounds.