Thinking Machines wants to build an AI that actually listens while it talks

Thinking Machines Lab unveiled a full-duplex interaction model that can process user input and generate output at the same time.

The startup said the system achieved a 0.40-second response time, aiming for more natural, phone-like conversation.

A limited research preview is expected in the coming months, with a broader release planned later this year.

If the claims hold up, the approach could set a new bar for real-time AI assistants and reduce turn-based friction.