An always-on assistant that listens to the room has one hard job: knowing when it isn't being asked.
On Thursday's show my co-host Nico Lafakis told this story about his dot (OpenAI's always-on assistant inside ChatGPT), with its voice on. He had a "full-blown conversation, questions involved, back and forth," in the room with his wife. The dot said nothing.
Afterward he asked how it knew to stay out, since it had said it couldn't voice print people (identify them by the sound of their voice). Its answer, as Nico tells it: "I know the way that you talk versus the way that [his wife] talks."
It worked from the words coming in, and "understood that somebody else was speaking because the language wasn't the same."
If you're deciding whether to leave an assistant listening, this is one account of how it decides when to speak up.
Watch the episode: https://valuefirststudios.com/media/episode/vf-ai-daily-2026-10-08
The next step