Mira Murati steps back into the spotlight, carefully
Murati’s first major interview in 18 months previewed Thinking Machines’ new “interaction models” and revisited OpenAI’s 2023 leadership chaos.
Intelligence analysis by GPT-5.4 Mini
Mira Murati used a rare public appearance to signal that Thinking Machines Lab is still moving, even if quietly. She described a new AI interface concept, reflected on OpenAI’s 2023 leadership crisis, and warned that too much power sits in too few hands.
Murati is showing a new kind of AI tool that may listen to speech, text, and video almost in real time, like a very fast helper that catches pauses and corrections. She also says AI companies need stronger rules so too much power does not sit in just a few people’s hands.
Analysis
A careful re-entry
Mira Murati, who was previously OpenAI’s CTO and later briefly its interim CEO, spoke publicly for the first time in roughly 18 months in a Bloomberg interview in San Francisco. TechCrunch frames the appearance as a deliberate but cautious move: Thinking Machines Lab has spent the past year and a half building in the background, raising capital, hiring researchers, and shipping one product, Tinker, an API for fine-tuning open-source AI models.
What Thinking Machines is teasing
Murati previewed what the company calls “interaction models.” She described them as a different kind of AI interface from the usual prompt-and-response setup. According to her description, the models are designed to process continuous streams of audio, text, and video in 200-millisecond intervals, aiming to capture things like interruptions, corrections, and pauses in near real time. She did not give a release date and made clear that this is only a first step.
OpenAI, governance, and the “blip”
Murati also revisited the chaotic week in November 2023 when OpenAI’s board fired Sam Altman and she became interim CEO, a period the company later nicknamed “the blip.” She said she felt clear about her choices at the time and believed the company would have “imploded” without her involvement, but she also said that in hindsight she would have pushed for more information, a better transition plan, and more transparency.
A recurring theme of the interview was governance. Murati argued that the industry has paid too much attention to the character of individual leaders and too little to structural checks. She said the bigger problem is the concentration of consequential decisions in too few hands, not just at OpenAI but across AI more broadly.
Competition and talent pressure
Asked about departures from Thinking Machines, Murati downplayed them, saying that building a frontier lab from scratch compresses ordinary organizational churn into a short period. She also acknowledged the pull of huge compensation packages but suggested money is not the whole story. Her broader message was restrained: the company wants attention, but not theatrics.
Key points
- Murati made her first major media appearance in about 18 months and kept the message tightly controlled.
- Thinking Machines has so far shipped Tinker, an API for fine-tuning open-source AI models.
- The company is previewing “interaction models” that process audio, text, and video in 200-millisecond intervals.
- Murati revisited OpenAI’s 2023 leadership crisis and said the company would have “imploded” without her involvement.
- She argued that AI governance matters more than the virtues of any single leader.
If Thinking Machines’ interaction models work as described, they could make AI feel more natural and responsive than today’s chat-style systems. The company’s emphasis on governance could also help push the industry toward stronger checks around powerful AI decisions.
The new interface idea is still early, and Murati offered no release date, so it may take time before it becomes a real product. The talent churn and concentration-of-power concerns she raised also suggest that even well-funded AI labs can face instability and governance risks.
