I wonder what a true voice IM clent would look like. The beauty of I’m that is lost in clients like Skype is it’s impossible to maintain several asyncronous near real-time conversations at once. You can hear this problem is you ever listen to a taxi radio. Perhaps when you focus different windows it could catch up on the discrete voice clips since the last time you were on that window. The whole chat UI could be a button you hold down while speaking, like a HAM radio.