Conceptually that was more or less what they were trying to do with Windows 8 - provide affordances optimized for each input method. For example, with touch you

Conceptually that was more or less what they were trying to do with Windows 8 - provide affordances optimized for each input method. For example, with touch you use swipe/flick gestures to pan back and forth, with mouse a scrollbar appears and you use that. With mouse some commands have small controls that appear when you hover over an item, with touch those controls are on a bar at the bottom of the screen. And so on. What made it different from what Ubuntu is doing (as I understand it) is a couple of additional problems they were trying to solve: First, you're probably thinking about "hybrid" devices in terms of having a tablet that you use with touch exclusively for X hours, and then attaching a mouse and keyboard, having it become a "PC" and using it with those inputs exclusively for X hours. In this usage model it's OK to have pretty stark changes between modes because you don't mode switch very often. However, they really wanted to enable a usage model where "touch plus mouse plus keyboard" became more like "mouse plus keyboard" in that you can constantly switch between inputs without really thinking about it, or even use them simultaneously. For this model to work whatever optimizations are done for different inputs need to be less intrusive and more localized - generally more about hiding/showing certain controls than completely changing the layout. Second, there was a feeling that even for large displays and mouse/keyboard the desktop windowing model felt too heavy and required too much manual window management, at least for some (more casual) use cases. So they were hoping the model developed for tablets could double as a solution to the "Windows feels like too much work" problem on the desktop, especially for home PCs used for entertainment and communication. To me the recent rumors sound like they're continuing the same general philosophy (of having a common UI with optimizations for each input method), but maybe deprioritizing the above two goals somewhat, which would leave them freer to make more drastic per-input-method optimizations.