When you pair program with someone, or assist someone, for it to have any point at all isn't the idea that you always edit in ONE place, and that only one person edits at a time? How will personA know what personB is doing otherwise, if personB's cursor can be anywhere?
Next, when personA and personB agree they have gotten somewhere and they launch their calculator app, web browser or whatever, how will both see what is going on? Is there support for also sharing a window/desktop? I mean, sharing my debugger state is cool, but what about the actual program window we are debugging? How can Jeff help me by introspecting variables in my program, if he can't see the program window?
The simpler problem of "sharing a screen while coding" seems to me to be a solved problem: you share your screen and input, like any remote desktop solution has done for ages. I get that this is significantly lower bandwidth and more tailored to coding, but in return it seems to not solve many of the fundamental problems of collaborative development.
I can see this tech being much more usable to e.g. develop cloud based IDE's, or remote debugging on servers, than "collaborative development"! Am I overly pessimistic or missing the point?