It isn't a browser, it's a windows shell replacement.
Full 3d doesn't add much value to the desktop experience and it isn't the direction we should be going. Instead I think we need to focus on make computers smarter, not prettier. We've basically plateaued on the prettiness factor. Sensor fusion, AI, and user customiziability are definitely the way to go.
Imagine having a camera/mic/IR sensor combo connected to your computer. Software can track user expression, posture, mood, etc and then based on the information it collects make an intelligent choice on which song you would most like to hear from your collection.
How about the computer tracking your frustration and learning what you really want. Based on expert systems the computer can ask questions and try to deduce what you want. If the frustation level gets to high it can simply switch over to a human connect at the software manufacturer.
Finally, different people have different needs and thus we need to make software that is modular to the point that one could drag and drop features from one package to another. Let's say a programmer writes a spell checker. I should be able to drag and drop that module into any application and have it work. This type of behavior requires standarized data types, which is what XML is good for.