Correction, it keeps a rolling loop going so that when the CPU hears the wake word it has enough audio buffered for voice recognition to have a clean starting p

Correction, it keeps a rolling loop going so that when the CPU hears the wake word it has enough audio buffered for voice recognition to have a clean starting point. If it only started recording when the wake word was processed as being heard there would be an added delay before you could start speaking a command. Might as well push a button and wait for a beep for that added delay.