For eighteen months, Mira Murati was the most prominent ghost in Silicon Valley. As the former CTO of OpenAI, she was the steady hand during the company’s most volatile public crisis, yet she spent the year following her departure in near-total silence. On Thursday, that changed. Sitting down with Bloomberg in San Francisco, Murati finally stepped back into the spotlight—not to rehash the past, but to signal that her new venture, Thinking Machines Lab, is ready to compete.

The Pivot to 'Interaction Models'

Murati didn’t come to the stage to offer a victory lap. Instead, she used the appearance to preview what she calls "interaction models." The current AI paradigm is defined by the turn-based, prompt-and-response dynamic—you ask, the model waits, then it answers. Murati argues this is fundamentally broken.

Thinking Machines is building models designed to process continuous streams of audio, text, and video in 200-millisecond intervals. The goal is to capture the texture of human communication: the interruptions, the mid-thought corrections, and the natural pauses that define how we actually think and speak. It is a technical bet that the future of AI isn't just smarter answers, but a more fluid, real-time presence. However, Murati was careful to frame this as a first step, declining to provide a release date or a finished product roadmap.

Governance Over Virtue

Inevitably, the conversation turned to the "blip"—the chaotic five-day stretch in November 2023 when OpenAI’s board fired Sam Altman and Murati was thrust into the interim CEO role. While she defended her actions as necessary to prevent the company from "imploding," she expressed regret over the lack of a transition plan.

More importantly, she used the moment to pivot to a broader critique of the industry. Murati argued that the current AI landscape suffers from a dangerous concentration of power. Her concern isn't just about the character of individual leaders, but the absence of structural checks. "Good people make bad calls," she noted. "Well-intentioned organizations drift." She suggested that the industry has spent too much time debating the virtue of its leaders and not enough time building the governance required to keep them in check.

The War for Talent

Building a frontier AI lab is a high-stakes game of musical chairs. Thinking Machines has seen its share of high-profile departures, a topic Murati has previously avoided. She downplayed the churn, suggesting that building a lab from scratch compresses years of organizational volatility into mere months.

While she acknowledged that the nine-figure compensation packages currently standard in the industry capture headlines, she argued they rarely tell the whole story. When asked about her competitive instincts, she offered a pointed contrast to the "move fast and break things" ethos of her peers: "When I wake up in the morning, I am not thinking about how to kill the competitor."

Key Takeaways

  • Interaction Models: Thinking Machines is moving away from prompt-based AI toward 200-millisecond, continuous-stream processing to mimic human communication.
  • Governance Critique: Murati argues that the AI industry is failing because it relies on the "virtue" of leaders rather than implementing necessary structural checks and balances.
  • Measured Growth: Despite high-profile departures and intense competition from OpenAI, Anthropic, and xAI, Murati is positioning her lab as a long-term project rather than a sprint to market.

What This Means for the Industry

Murati’s return is a reminder that the next phase of the AI arms race will be defined by more than just raw compute. As the market matures, the focus is shifting toward how these models integrate into the messy, real-time flow of human life.

Whether Thinking Machines can deliver on its promise of fluid interaction remains to be seen. But by re-entering the public conversation, Murati has signaled that the era of quiet building is over. The industry is now watching to see if her structural concerns about governance will gain traction, or if the race for dominance will continue to outpace the need for oversight. The next decision point for her lab will be its first public product release; until then, the industry will be left to wonder if her vision of a more "human" AI can survive the pressures of the market.