The TL;DR
Former OpenAI CTO Mira Murati made her first public appearance in 18 months, previewing Think Machines Lab’s “collaborative models” and arguing that the AI industry lacks structural control. He also talked about the researcher’s departure and hinted at Altman’s 2023 shooting.
For someone who helped ship ChatGPT, DALL-E, and Codex, Mira Murati is surprisingly quiet. On Thursday, he broke his silence. Sitting down with Bloomberg’s Emily Chang in San Francisco, the CEO of Thinking Machines Lab made his first media appearance in nearly 18 months, a carefully controlled re-entry into a conversation that has moved too fast without him.
The timing was not accidental. Think Machines spent that year and a half raising $2 billion, acquiring a gigawatt of Nvidia Vera Rubin compute, shipping one product, and losing an alarming number of researchers it hired to build the next. The AI landscape Murati left behind when he left OpenAI in September 2024 looks nothing like the one he re-entered on Thursday.
Product: interactive models
Murati used the interface to preview what Thinking Machines calls “interaction models,” a very different type of AI interface. Instead of the information-and-response format that defines most AI products, the company’s models are designed to process continuous streams of audio, text, and video at 200-millisecond intervals.
TNW City Coworking Space – Where your best work happens
A workplace designed for growth, collaboration, and endless networking opportunities at the heart of technology.
The pitch is that these models can continue the texture of human communication: interruptions, mid-thought corrections, pauses. The name of the technology is “full duplex,” and the company says its TML-Interaction-Small model responds in 0.40 seconds, about the speed of a natural conversation. It goes along with the founding thesis of the Thinking Machines that powerful AI requires close human cooperation, not less of it.
Murati made sure to do this as a first step. He declined to put a release date on anything, and put the work aside for Tinker, the company’s API for fine-tuning open source models, which was launched in October 2025 and is its only shipping product.
Walking
Chang pressed Murati on what has become a seemingly silent problem for the company: a series of high-profile departures. Founder and CTO Barret Zoph, founder Luke Metz, and founding team member Sam Schoenholz all returned to OpenAI in January. Five founding members went to Meta, reportedly with nine-figure compensation.
Murati underestimated the exit. Building a frontier AI lab from scratch compresses years of typical organizational stagnation into months, he said. He acknowledged that nine-figure packages are now common in AI talent war capture concepts, but suggested compensation is rarely the whole story.
“When I wake up in the morning, I don’t think about how I will kill my competitor,” he said, drawing laughter from the audience. The line was disarming, but the reality of the competition is clear. OpenAI is everywhere. Anthropic has raised $30 billion and is reported to have attracted offers from investors worth $800 billion. Elon Musk’s xAI folded into SpaceX ahead of record IPO. In that case, staying silent has a cost.
Altman’s shooting, revisited
Chang asked about the episode that first made Murati a public figure: the five-day upheaval in November 2023 when OpenAI’s board fired Sam Altman and Murati became interim CEO. Within OpenAI, this phenomenon was called a “blip.”
Murati said that he feels clear about his decisions at all times, that protecting the mission and the team is the thread that makes the choice seem obvious as the situation seems to be falling apart from the outside. He said the company would have “exploded” without his involvement at the time. But he acknowledged that clarity of purpose is not the same as clarity of results, and said he would have pushed harder for more information, a better reform plan, and more transparency.
Asked if he still trusted Altman, he backed off. What he offered instead was very interesting: a broad debate about the concentration of important decisions in the hands of too few, not only in OpenAI but in the industry as a whole. He said his concern is less about the personality of any leader but about the lack of structural checks. Good people make bad calls. Well-intentioned organizations are drifting.
A difficult question
For the future of AI more broadly, Murati pushed back against both a dystopian and a utopian framework. He argued that there is no predetermined outcome. The time we are in right now will determine how things go.
But he returned, more than once, to a theme that links his criticism of management to his brand’s philosophy: if people take their hands off the wheel too soon, the future will look very different, and not better. It’s a position that sits comfortably with his company’s thesis about human interactions with AI. Whether it can survive dealing with a market that rewards speed, scale, and tens of billions of dollars with caution is a question Murati did not answer Thursday.
He doesn’t need to answer it yet. But with a single product launch, a shrinking team at the top, and a growing number of competitors by the week, the window for peace is closing.
