It's obviously always listening since it has to respond to the trigger word. I assume that it's got a little buffer, and when it detects the trigger word it upl

It's obviously always listening since it has to respond to the trigger word. I assume that it's got a little buffer, and when it detects the trigger word it uploads the spoken text plus a little buffer of time before the trigger word was spoken.