Mira Murati Emerges to Preview ”˜Interaction Models’ and Address

What happened

Mira Murati has stepped back into the spotlight for the first major time in 18 months, using a high-profile interview to introduce Thinking Machines Lab's work on what she calls "interaction models" and to revisit the OpenAI leadership crisis that briefly made her interim chief executive in 2023. The appearance marks a shift in posture for her new company, which is moving from relative opacity toward active public positioning in a crowded frontier AI market.

Murati's comments matter because she is one of the few executives who has stood at the center of both the most consequential governance rupture in modern AI and the creation of a new lab intended to compete with the companies she once helped lead. When she talks about product design and power concentration at the same time, it is difficult to separate the two.

What's new in this update

Thinking Machines is now previewing "interaction models" that are designed to process continuous streams of text, audio, and video in roughly 200-millisecond intervals. Murati's description suggests a move away from turn-based prompt-response interaction toward systems that can track live conversational flow, interruptions, timing, and real-time context more naturally.

That is a notable product direction because most mainstream AI systems still behave like highly capable turn-taking tools. Murati appears to be arguing that the next interface leap will not come only from smarter answers, but from models that can engage with human communication as an unfolding stream rather than a sequence of isolated prompts.

Key details

Thinking Machines has not announced a public release date, but Murati's remarks make clear that the company wants to define a more real-time, interaction-centric layer of AI computing. The company has so far been known publicly mainly through Tinker, an API for working with open-source models, but the new comments suggest a much broader ambition.

Murati also used the interview to revisit OpenAI's November 2023 board crisis, saying she helped prevent the organization from imploding while also acknowledging regret over how opaque and chaotic the process became. That dual message is revealing: she is simultaneously claiming operational credibility and highlighting a structural failure in how frontier AI labs are governed.

Several themes now define her public positioning:

• Interaction quality may become as important as raw model intelligence.
• Real-time AI interfaces require different technical design than chatbot products.
• Frontier labs still lack adequate governance checks and transparency.
• The next generation of AI companies may try to differentiate on both interface and institutional structure.

Background and context

Murati's departure from OpenAI and subsequent launch of Thinking Machines happened against a backdrop of extraordinary concentration in AI power. A small number of labs now control the most capable models, the biggest capital flows, and much of the public narrative about how the technology should evolve. Murati has now made clear that she sees this concentration as a problem, not just a market reality.

That concern gives additional meaning to the company's technical direction. If interaction models are meant to be more embedded and continuous, then governance questions become even more important. A system that is always listening, interpreting, and responding in real time could be much more powerful, but also much more sensitive in terms of control, trust, and failure modes.

What to watch next

The next real test is whether Thinking Machines can show a working product that makes the interaction-model idea concrete. Murati's stature guarantees attention, but it will take demonstrations and developer traction to prove that the company is offering something meaningfully distinct from OpenAI, Anthropic, Google, or Meta.

Why this matters

This matters because Murati is trying to define a new AI company around two ambitious claims at once: that the interface layer of AI is about to evolve beyond turn-based chat, and that the governance structure of frontier labs needs deeper reform. If Thinking Machines can connect those claims in product form, it could become an important counterweight in the next phase of the industry.

Reader context

This story belongs to Northstar Herald's OpenAI and Generative AI coverage, with related entities including Mira Murati, Thinking Machines Lab, Interaction Models, Corporate Governance. The report is based on TechCrunch AI source material.

Related coverage

• More ai coverage
• OpenAI updates
• Earlier coverage: Meta Deploys Rapid AI Data Centers in Ohio Using Weatherproof Tents