Preface

Mira Murati has long been a powerful but reserved presence in the AI world. Formerly OpenAI’s chief technology officer and now CEO of Thinking Machines Lab, she rarely courts headlines. Her recent, high-profile interview with Bloomberg in San Francisco — her first major media appearance in about 18 months — offered a careful reintroduction. The conversation served dual purposes: to remind observers that Thinking Machines is active and to outline the early thinking behind a new class of AI interfaces. In short, Murati used the occasion to re-emerge deliberately, signaling ambition without overpromising and emphasizing governance and responsibility as central concerns.

Lazy bag

Key takeaways: Murati reappeared publicly after a long hiatus to highlight Thinking Machines’ work on so-called interaction models, which aim to process continuous multimodal input in near–real time. She reflected on the tumultuous week at OpenAI in November 2023, framed organizational challenges as governance problems rather than solely personality-driven ones, and downplayed recent staff departures as part of early-stage organizational volatility. Overall, Murati struck a tone of cautious optimism while emphasizing structural checks and careful product development.

Main Body

Mira Murati’s return to the media spotlight was notable less for dramatic revelations than for the signals embedded in her restraint. As a former CTO of OpenAI, she had been a key architect of the company’s technological trajectory but seldom its public face. Since founding Thinking Machines Lab, her public appearances have been sparse. That makes a sit-down interview with Bloomberg — her first sustained interaction with a major outlet in roughly a year and a half — meaningful not for bombshells but for what it reveals about priorities and positioning.

Thinking Machines has spent much of the past 18 months operating quietly: raising funds, assembling research talent, and shipping an initial product called Tinker, an API intended to make it easier to fine-tune open-source AI models. In a market that has increasingly rewarded visibility, the company’s low-profile approach risks fading into the background unless it periodically asserts its existence. Murati’s Bloomberg appearance was, in part, exactly that: a measured reminder to the industry, investors, and potential customers that Thinking Machines remains active and evolving.

At the heart of the interview was Murati’s description of what she calls interaction models. These are positioned as a departure from today’s dominant, turn-based prompt-and-response paradigm. Rather than discrete exchanges, the firm’s models are built to process continuous streams of audio, text, and video in roughly 200-millisecond increments. The aim is to capture the nuances of human conversation — interruptions, mid-thought corrections, contemplative pauses — and to approach a more fluid, conversational tempo that mirrors real-time human interaction. Importantly, Murati emphasized that this is an early-stage concept: a first step that still needs refinement and for which no concrete release timeline was offered.

Murati also confronted topics that have followed her into the public record, most notably the chaotic five-day period in November 2023 when OpenAI’s board removed Sam Altman and Murati served briefly as interim CEO. Internally dubbed "the blip," those days were chaotic and consequential. Murati described her decision-making during that stretch as guided by a consistent priority: protecting the organization’s mission and team. She argued that her involvement helped prevent immediate organizational collapse, while conceding that hindsight revealed areas where she would have pressed for clearer information, a more robust transition plan, and greater transparency.

When asked whether events ultimately worked out well — and whether she still trusts Altman — Murati declined to offer a direct verdict. Instead she broadened the conversation to a systemic concern she repeatedly returned to: the concentration of high-stakes authority in relatively few hands across the AI sector. Her worry was not merely about individual character or competence, but about institutional structures. Good people can still make damaging decisions; organizations can drift; and relying on individual virtue without governance safeguards is a fragile approach. In Murati’s view, the industry has emphasized the moral qualities of leaders while underinvesting in mechanisms that ensure accountability and distribute decision-making.

The interviewer also raised the departures of several prominent researchers from Thinking Machines, a subject Murati has seldom discussed publicly. She framed those exits within the context of early-stage lab-building, where turbulence that might take years in a mature organization can happen in months. She acknowledged the magnetic pull of large compensation packages now common in AI recruiting — nine-figure offers have become a headline-grabbing feature of the talent marketplace — but suggested pay alone doesn’t fully explain why people move. She downplayed narratives of acrimony, emphasizing instead the compressed timelines and shifting priorities that accompany the formation of a new research organization.

Throughout the interview, Murati maintained a careful tone when addressing broader concerns about AI’s societal impacts. She rejected simple arc narratives of inevitable utopia or inevitable catastrophe. Instead she argued that the present moment is decisive: policy, governance, corporate behavior, and research priorities set now will strongly influence long-term outcomes. She repeatedly warned against premature abdication of human oversight — "taking hands off the wheel" too early, in her phrasing. Doing so, she cautioned, would make a less desirable future more likely.

Murati’s posture combined ambition with humility. On product strategy, she previewed a vision that seeks richer, more natural human–machine interaction. On organizational and industry issues, she emphasized the need for stronger structures and transparency to prevent concentration of risk. And on the human consequences of AI, she maintained a balanced skepticism about deterministic claims while urging active stewardship.

Her reemergence illustrates a common tension for early-stage AI leaders: the need to attract attention and resources without overstating progress or promising undue certainty. In a field where competitors and narratives can reshape markets quickly, the option to remain entirely quiet grows less viable over time. Yet visibility also brings scrutiny. Murati’s interview demonstrates a strategy of incremental disclosure: share enough to reassert relevance, frame ideas as exploratory rather than final, and repeatedly emphasize governance as a parallel priority to innovation.

In sum, Murati’s carefully calibrated return to the public stage was not a full unveiling but a strategic reminder. Thinking Machines continues to develop new interface paradigms and to recruit talent amid a fraught and competitive landscape. At the same time, Murati wants to push the conversation beyond product features to the institutional norms and governance frameworks that will shape how AI’s promises and risks play out. Her message: progress matters, but so do the rules and structures that guide it.

Key Insights Table