Mira Murati is back on the record. The Thinking Machines Lab CEO and former OpenAI CTO sat down with Bloomberg in San Francisco on Thursday for her first major media appearance in roughly 18 months, according to TechCrunch AI. She didn’t say much. But what she did say tells you where her company is headed and how she’s reading the moment.
What stands out here is the timing. Thinking Machines has spent close to a year and a half heads down: raising money, hiring researchers, and shipping a single product called Tinker, an API for fine-tuning open-source models. Meanwhile the competition got louder. OpenAI lives in the news cycle. Anthropic’s momentum is the only thing people want to talk about. And xAI just got folded into SpaceX ahead of an expected blockbuster offering. In that crowd, staying quiet stops paying off. At some point you have to make noise just to remind the market you’re still in the game.
The product: ‘interaction models’
Murati used the interview to preview what Thinking Machines calls interaction models, which she described as a fundamentally different kind of AI interface.
Here’s the idea, as she laid it out to interviewer Emily Chang:
- Instead of the turn-based, prompt-and-response pattern that defines most AI products today, the models process continuous streams of audio, text, and video.
- They work in 200-millisecond intervals, close to real time.
- The goal is to pick up the texture of human communication: interruptions, mid-thought corrections, even pauses to think.
This is significant because the prompt-response loop is the dominant interface for nearly every chatbot you use. A model that reads conversation as a live stream instead of discrete turns would feel less like typing into a box and more like talking to someone. Murati was careful, though. She framed it as a first step, not a finished product, and refused to put a release date on anything.
Looking back at ‘the blip’
Chang also pulled Murati back to November 2023, the chaotic five days when OpenAI’s board fired Sam Altman and she stepped in as interim CEO. Inside the company they called it the blip.
Murati said she felt clear in each moment, with protecting the mission and the team as her through-line. She said OpenAI would have imploded without her involvement during that stretch. But she drew a sharp line between clarity of intent and clarity about consequences. In hindsight, she said, she would have pushed harder for more information, a better transition plan, and more transparency.
Asked whether she still trusts Altman, she sidestepped. Instead she steered toward a bigger worry she returned to repeatedly: too many consequential decisions sit in too few hands, not just at OpenAI but across the industry. Her concern isn’t mainly about any single leader’s character. It’s about the absence of structural checks. Good people make bad calls. Well-meaning organizations drift. The industry has paid too much attention to virtue, she suggested, and not enough to governance.
On talent, jobs, and the road ahead
Murati downplayed the recent departures of high-profile researchers from Thinking Machines. Building a frontier lab from scratch, she said, compresses years of normal organizational turbulence into months. She acknowledged the nine-figure pay packages now standard in the talent war grab attention, but argued money usually isn’t the whole story. On her own drive, to some audience laughter: “When I wake up in the morning, I am not thinking about how to kill the competitor.”
Asked about job displacement and darker risks, she rejected both inevitable utopia and inevitable dystopia. The period we’re in now, she argued, is the one that decides which way it goes. Her warning, repeated more than once: if humans take their hands off the wheel too soon, the future will look very different, and not better.
For more on Murati’s return and the interaction models reveal, see the full report at TechCrunch AI.