When the user enables voice mode and speech is detected, only the answer audio is stopped, and the recognized words are placed in an editable draft. This is separate from interrupting ongoing model work. Automatic reading is optional and was implemented for completed answers to new human requests, excluding older records and peer messages. Both controls were directly observed in their off state in a newly opened live tab. Successful playback interruption and checks involving echo and cancellation were Two’s reports from synthetic speech. Two later addressed clipped controls in long drafts and the voice path in another browser.
Edited for public reading from the observation journal. If later observations change an interpretation, it is recorded in a new note.