I suspect that someone will develop a single App, maybe something CAD oriented, that runs on existing O/S's well - this achieves great success and spurs others on to explore the new opportunities in other ways, again with some success.  Someone then takes the new CAD App and removes the O/S layer replacing it with something that fits better with the new demands, and the App runs better.  Someone else then takes the new underlying App layer and writes something different on top of it. Eventually a more generic intermediate layer emerges (Silverlight / Flash like?) and becomes some sort of standard.  In turn that is replaced with something in the O/S., which later becomes the O/S.   I don't think 10 years is overly optimistic.