Thinking Machines, the AI startup founded by former OpenAI CTO Mira Murati, just announced what it's actually building: interaction models. Think of them as AI that works more like a real conversation partner instead of waiting for you to finish typing before it starts thinking.
Here's the core idea. Current AI models work in a single thread. You type or speak, hit send, then the model processes everything and responds. Thinking Machines wants to change that by building models that continuously take in audio, video, and text at the same time, processing and responding in real time.
The company describes it as letting people "collaborate with AI the way we naturally collaborate with each other." That means the AI would perceive what you're doing and how you're doing it as it happens, not after you've finished your entire thought.
This matters because the turn-based nature of current AI tools creates friction. You can't interrupt, clarify mid-sentence, or have the kind of dynamic back-and-forth that makes human collaboration effective. If Thinking Machines pulls this off, it could make AI assistants feel less like chatbots and more like actual working partners.
Murati left OpenAI in September 2024 and raised funding for Thinking Machines shortly after. The company hasn't shared timelines or technical details about how they'll build these interaction models, but the ambition is clear: make AI that thinks and responds the way humans actually communicate.
