Mira Murati spent six years inside OpenAI, four of them close to the centre of the most consequential product launch in modern software. Then she walked out. Less than a year later she was running the most talked-about startup in artificial intelligence.
That arc — from an engineering graduate out of Dartmouth to interim chief executive of OpenAI to founder of Thinking Machines Lab — says a lot about where AI is heading, and about who gets to decide how it gets there.
From Vlorë to Dartmouth: An Unlikely Route Into Tech
Murati was born in 1988 in Vlorë, a port city on Albania’s Adriatic coast, in the final years of an isolated communist regime. Her father taught mathematics. When the government fell in the early 1990s, the country opened up, and so did her options.
She won a scholarship to Pearson College, a United World College on Vancouver Island, and from there went to Dartmouth College, graduating in 2012 with a degree in mechanical engineering. Her early career reads like a tour of ambitious engineering projects rather than a straight line into software:
- Goldman Sachs — a brief internship she has since described as a poor fit
- Zodiac Aerospace — work on aircraft interior systems
- Tesla — engineering contributions to the Model X
- Leap Motion — vice president of product at a startup promising to replace the mouse with hand gestures
Leap Motion never became the next big input device, but it taught her something that shows up in everything she has done since: how people actually interact with a machine matters as much as what the machine can do.
The OpenAI Years: DALL·E, ChatGPT, and Five Very Public Days
Murati joined OpenAI in 2018 as vice president of applied AI and partnerships, back when the lab was still a research outfit with a strange name and no revenue to speak of. She became chief technology officer in 2022, and then the products started landing in quick succession — DALL·E 2, ChatGPT on 30 November 2022, GPT-4 in March 2023, and the video model Sora in early 2024.
She was the public face of much of it, the person who stood on stage and explained what these systems could and couldn’t do. Colleagues described a manager who could translate between research and product teams that barely spoke the same language.
The 2023 boardroom crisis
Then came the weekend that turned her into a household name in tech circles. On a Friday in November 2023, OpenAI’s board fired Sam Altman and named Murati interim CEO. Within days, most of the company had signed a letter demanding his return, and Murati’s name was on it. Five days later Altman was back and she was CTO again.
For more than a year, what happened inside those five days stayed largely private. Her later testimony added real detail, including how she weighed the information the board gave her against what she knew herself. If you want the granular version, the account of how Murati’s deposition pulled back the curtain on Altman’s ouster is the place to start.
Why She Left in September 2024
Murati resigned from OpenAI on 25 September 2024, announcing it in a short post that thanked the team and said she was stepping away “to create the time and space to do my own exploration.” She left alongside a wave of senior departures that reshaped the company’s leadership bench.
What followed the resignation was unusually quiet for someone with her profile. No podcast tour, no manifesto. Just a period of silence that ended in February 2025 with a single announcement.
Thinking Machines Lab: The Company She Built Instead
Thinking Machines Lab launched with roughly thirty employees, a large share of them former OpenAI colleagues, and a name borrowed from a 1980s MIT spinout. The pitch was deliberately unglamorous compared with the race to build ever-larger models: make AI systems more useful, more customisable, and easier for ordinary organisations to shape around their own needs.
The money and the structure
In July 2025 the company closed a seed round reported at $2 billion, at a valuation around $12 billion — by most measures the largest first funding round in Silicon Valley history. Rather than a single flagship chatbot, the lab has leaned into tools that let customers adapt models themselves, including a fine-tuning service aimed at teams that want control over their own stack. For a closer look at the product direction, this breakdown of what Murati’s AI company is actually building fills in the technical detail.
What “keeping humans in the loop” means in practice
The phrase sounds like corporate boilerplate until you look at how she has applied it. Murati has consistently argued that the interesting future isn’t one where a model answers and a person receives, but where the two work on a problem together — with the human setting direction and the model handling scale. She has framed that as an AI designed to keep humans in the loop rather than a system that quietly replaces them.
It’s a positioning that cuts against the way most labs market themselves, and it has practical consequences for how her products are built — more adjustable defaults, more room for people to shape behaviour after training, less reliance on one enormous model doing everything. The wider argument, that TML upends how humans work with AI, is worth reading alongside it.
Under Oath: The Courtroom Chapter
Murati’s name surfaced again in 2025, this time in a federal courtroom. As part of the litigation between Elon Musk and OpenAI, she sat for a deposition that was later unsealed, and her answers were notably careful. She described a working relationship with Sam Altman that had limits, and testified that she could not always take his statements at face value.
Those remarks carried weight precisely because she had been in the room for the decisions in question. You can read the substance of her testimony on why Murati told the court she couldn’t trust Altman’s words. For anyone following OpenAI’s governance story, it is the most detailed inside account yet.
The Race She’s Now Running In
Thinking Machines Lab competes with companies that are far larger and far better funded, from Anthropic and Google DeepMind to Meta’s superintelligence group and Musk’s xAI. The frontier keeps moving, too — OpenAI’s next generation of models, including the generational leap promised with GPT-6 Astra, is expected to reset expectations again.
Murati’s bet is that scale alone won’t decide the winner. If customisation, reliability, and human oversight turn out to matter as much as raw benchmark scores, a leaner lab with a clear thesis can hold its ground. It’s the same instinct that ran through her work at Tesla and Leap Motion: the interface is the product.
What to Watch Next
Murati is now the rare AI leader who has been inside the dominant lab, briefly led it, testified about it, and is building a competitor to it. A few things are worth tracking from here:
- Whether TML ships anything consumer-facing. So far the emphasis has been on developer and enterprise tooling, not a headline chatbot.
- How the team holds together. Founding teams in AI are volatile, and losing senior researchers to rivals is the sector’s most common failure mode.
- Whether “human in the loop” survives contact with the market. Customisation costs money and slows deployment. Buyers will decide if the trade is worth it.
- What she says publicly. She has been selective with interviews, and each appearance tends to move the conversation about where the industry is heading.
For a founder who once described herself as a mechanical engineer who wandered into AI, she has ended up at the centre of nearly every important argument the field is having — about safety, about governance, about who controls the most powerful technology of the decade. Whatever Thinking Machines Lab becomes, that position isn’t going away.