Mira Murati steps back into the spotlight, carefully

2026-06-05 · Source: AI News & Artificial Intelligence | TechCrunch · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Fundamental Awareness, short

Summary

Mira Murati, CTO of OpenAI and CEO of Thinking Machines Lab, made her first major media appearance in roughly 18 months with Bloomberg, signaling a more public phase for her company. Thinking Machines Lab, after operating quietly for a year and a half, has raised capital, hired researchers, and shipped Tinker, an API for fine-tuning open-source AI models. Murati previewed "interaction models," a new AI interface designed to process continuous audio, text, and video streams in 200-millisecond intervals, aiming for real-time human communication understanding. She also reflected on the November 2023 OpenAI "blip," stating her decisions protected the mission and team, preventing implosion. Murati expressed concern about the concentration of consequential decisions and lack of structural checks in the AI industry, emphasizing governance over individual virtue. She also addressed recent high-profile researcher departures, attributing them to organizational volatility and compensation.

Key takeaway

For AI product managers evaluating next-generation interfaces, Murati's "interaction models" suggest a shift from turn-based prompting to continuous, real-time processing of multimodal input. You should explore how your current AI systems could adapt to process audio, text, and video streams in short intervals to better capture human communication nuances. Prioritize governance structures in your AI development to mitigate risks from concentrated decision-making.

Key insights

Thinking Machines Lab is developing "interaction models" for real-time, continuous human-AI communication beyond turn-based prompts.

Principles

• AI industry governance requires structural checks, not just individual virtue.
• Frontier AI lab development compresses organizational volatility into months.

Method

Thinking Machines' "interaction models" process continuous audio, text, and video streams in 200-millisecond intervals to understand human communication nuances.

In practice

• Consider fine-tuning open-source AI models using an API like Tinker.
• Evaluate AI interfaces for continuous stream processing capabilities.

Topics

• Thinking Machines Lab
• Interaction Models
• AI Governance
• AI Interfaces
• Tinker API
• OpenAI Leadership

Best for: CTO, VP of Engineering/Data, AI Architect, Director of AI/ML, AI Product Manager, Investor

Related on AIssential

• Here’s what Mira Murati’s AI company is up to — Interaction models aim to enable real-time, multimodal human-AI collaboration by continuously processing diverse inputs.
• Towards AI That Can Actually Interact — Real-time, continuous interaction models represent a fundamental shift from turn-based AI, enhancing human-AI collaboration.
• Thinking Machines Just Announced More Human Like AI — Thinking Machines Lab introduces "Interaction Models" for real-time, human-like AI collaboration via simultaneous input/output.
• New AI Model SHOCKED Me — Thinking Machines Labs' new model offers real-time, multimodal, and context-aware conversational AI.
• [NOTICE] This feed is no longer maintained — Native interaction models enable real-time, multimodal human-AI collaboration by integrating interactivity directly into the AI's core architecture.
• Thinking Machines shows off preview of near-realtime AI voice and video conversation with new 'interaction models' — Thinking Machines' interaction models enable real-time, full-duplex human-AI communication by processing multimodal inputs and outputs simultaneously.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI News & Artificial Intelligence | TechCrunch.