My initial thoughts are that it does what is says on the tin and is pretty good at doing it to boot.

But ....

To me this kind of interaction is crying out for its own Operating System UI - which I don't believe is a logical progression from where we are today (with Windows 8 or Unix derivative UI's).

I have an idea what this might be but have struggled to put it down on paper:

Watch the video in the link above and then the following will maybe make more sense

To me it seems like our hands and arms will become 'virtually extended' such that they exist behind the screen within a 3 dimensional (stereoscopic) plain. 

A new set of Applications need to be developed that exist within this 3d plain and are designed for interaction by our (virtual) hands and fingers. (Noting that most interaction will probably take place with minimal movement by our fingers).

I believe that this will feel very natural, perhaps in the same way that a car becomes an extension of our bodies when we drive?

I guess audio commands/feedback and some kind of tactile feedback to our fingers might also play a part?

10 years time?

And before someone says 'Minority Report' I would say yes - but that was an Application designed to locate and view documents (if memory serves) - but there must be plenty of other functional uses for this style of interface?

This said, I'm kind of worried about how well this would work for the general public, looks like there's going to be a "palm rejection" problem on steroids here. Either users get used to move their hands off the sensing area, or the UI needs to be very picky (or very smart) about intentional and valid gestures. Either way I expect it will need some training.

With a little luck, the coolness of "casting spells" to your PC will be a big enough incentive.

As for the audio commands and feedback, the technology never worked too well for PCs; one thing is to bark commands to an xbox in your living room, another is to have to say what you are doing in public. I doubt that's ever going to change, until they nail subvocalization.

Once that happens, we'll have proven Clarke's third law again: "any sufficiently advanced technology is indistinguishable from magic".

I'm not too sure about this whole "moving 3D objects around with your hands" thing in a general purpose OS/application, though. For CAD users, sure, but for others it just seems like too complicated a solution.

Bas is right though; it may be too much for users to handle for general purpose uses. Just look at people's reaction to the changes made to Windows for W8 -- not great. Then again people seems to take to the Kinect pretty quickly.

At the base of the OS, the input itself has to respect the degrees of freedom allowed by this kind of UI.  I mean, things like fonts and writing will have to be changed.  I think that natural language will evolve in huge leaps, as we abandon keyboard input as the primary input and a new generation of calligraphy emerges in 3D space.  People will be able to rapidly create 3D models, like manipulating clay, to express thoughts and ideas.  They will be able to animate it and make it interactive.  Text is not going away, but I think that a new world of media as a language will emerge.  That has to be baked into the OS.  A new kind of shell that makes 3D models first class.

EDIT:   I also think that technology like Photosynth will play a huge role, as it will allow hyperlinking images and 3D modeling.  This has already been demonstrated by Bing and Photosynth, i.e. an image in a picture links to a 3D model made from pictures of the same source.  So, imagine if I model something with some basic 2D outline, or even 3D virtual clay, like the shape of house I want to search for on Bing maps, and when I find it, I can create a hyperlink to related shapes.