I'm surprised yours is the only comment I saw pointing out the incremental approach to this "living room" goal Valve has been using.

The steam box will obviously require some OS to run, and it seems perfectly reasonable to ship the OS, get feedback for it and it's features (i.e. streaming), and have it production-tested once the inevitable steam box is ready. Why take the risk of shipping all-new-everything when you can do it in smaller steps?