I assume the naive part is that the keyboard polling interface probably doesn't allow that much input ? Otherwise yes, we do this kind of approach so much irl.

I assume the naive part is that the keyboard polling interface probably doesn't allow that much input ? Otherwise yes, we do this kind of approach so much irl. The most useful kind is the Switch it's button actuator, that will just push a button when making it "smart" is just impossible or not worth it. The less useful kind is what the Rabbit R1 was trying to do, with website interaction automation in a full VM because each service wouldn't provide an API.