Noticing a common theme... hardware that talks to the cloud has a >1sec response time, where hardware that runs locally responds instantly