A recorded call
The sentence that stops
He names the tablet, blocks for six seconds, and stops before the last word. A fragment cannot be relayed verbatim and is not a half-said name, so for the first time in these three calls both models run. Computer voices; everything that listens and speaks for Robert was live.
The call, moment by moment
Press play to hear the call
About 38 seconds. Sound on. Or step through it with the arrows.
Robert's screen
Ready when you are.
What is actually happening
- Silence protected
- 0.0s
- Times the line was held
- 0
- Inventions blocked
- 0
- Relayed to the other person
- 0
Who may speakCurrent state: Waiting
Listening twice
Same microphone, two AssemblyAI streams. The fast ear is set up like an ordinary voice agent.
Patient ear
max_accuracy, waits up to 9s
- Waiting for speech.
turns ended: 0
Fast ear
min_latency, default pauses
- Waiting for speech.
turns ended: 0
Pauses appear here once a sentence is finished.
Two AIs argue. A rule decides.
When the caller finishes a sentence, the decision about what to say for them appears here.
What the other person heard
The call has not started.
The record every call leaves behind: each line approved for the agent to say, whether it was spoken or cut off, every question the caller was asked and what went out as a result, and the post-call audit's verdict on each relay. It re-checks the guarantee that the voice said no word the rules did not approve.