Thinking Machines unveils real-time 'interaction models'
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Thinking Machines announced 'interaction models' that continuously process audio, video, and text, responding in real time rather than waiting for users to finish typing or speaking.
- The company framed today's AI as a 'bandwidth bottleneck,' arguing current models freeze perception while generating — comparing the experience to 'trying to resolve a crucial disagreement over email rather than in person.'
- Demos shared by Thinking Machines included listening for animal mentions in a story, real-time speech translation, and alerting a user when they're slouching.
- Thinking Machines plans a 'limited research preview' in the coming months, with a wider release 'later this year' — but the models aren't publicly testable yet.
- The lab was founded by Mira Murati in February 2025 after she left OpenAI, and has since lost key members to Meta and back to OpenAI.
Why it matters: For AI competitors, Thinking Machines is staking a distinct technical bet — continuous multimodal perception rather than the turn-based pattern every major lab currently ships. The limited preview is months away and the company has already seen talent drain to Meta and OpenAI, making execution speed the real test. For users, the pitch is interfaces that adapt to human behavior (catching your slouch, translating mid-sentence) instead of forcing rigid prompt-and-response.
Ask SkimNews
