The MIT wearable demo from a few years ago which used a similar concept to project an interface in the real world was incredibly compelling, but mostly because it assumed near flawless real world AI object recognition, along with flawless projection onto said items. They'll need to demonstrate this on this particular device, before this becomes remotely interesting. Yes, it's a "detail", but I think for a lot of this kind of tech, demonstrating just how DEEPLY you can go into the interactions is sort of the whole point if they are thinking of replacing the kinds of devices that we depend on.