Thinking Machines Lab has revealed its first Interaction Model after 18 months of development, led by Mira Murati, the former OpenAI chief technology officer (CTO). The project aims not simply to produce a more capable chatbot but to break away from turn-based AI interactions.
How this model differs
Most current AI products still confine user interaction to a prompt box: a user issues an instruction, the model replies, and the user inspects and corrects the output. The Thinking Machines Interaction Model is designed to listen and watch continuously, to interrupt and respond while an interaction is ongoing rather than waiting for discrete turns.
Key features shown include:
- 200 millisecond micro-turns to enable rapid, near-real-time responsiveness.
- Live audio and video handling, enabling multimodal interaction beyond text.
- Real-time search and UI generation to produce context-aware outputs quickly.
- A second background model that performs deeper, longer-running work without interrupting the live interaction.
Why it matters
If the approach proves practical, the prompt box could become a temporary interface for certain tasks rather than the central control point. Agents could evolve from ticket-system–style workflows into live collaboration layers where users steer, interrupt, and reshape execution in real time.
Limits and next steps
The announcement outlines the design and demonstrated capabilities but does not yet resolve questions around broad applicability, scaling, or real-world behavior in diverse user environments. Further practical testing and future updates from Thinking Machines Lab will be needed to assess production readiness and rollout plans.
