AI voice chats still feel awkward because assistants don’t know when to talk

Thinking Machines Lab unveiled TML-Interaction-Small, a research-preview model that listens and speaks at the same time.

The company says it can respond in about 0.40 seconds, aiming for more natural voice conversations.

Limited access is planned in the coming months, suggesting a cautious rollout.

If it performs well, the model could reduce awkward pauses and make AI assistants feel more human.