I'm a designer on windows 10 and I wanted to comment on your last thought. Take it with a grain of salt but these types of ideas are what we are working on every day consistently attempting to improve or refine. The following are my own thoughts:
I agree that the desktop interface is stagnating - but the WIMP methodology is still bar none for discoverability and learnability, and if you're a company making money from the enterprise market, you will always be torn between desiring and pushing innovation and maintaining consistency.
However one thing I think we are doing now that is very interesting is attempting to reconcile the world of WIMP (desktop) with the FIM paradigm (finger icon menu - I made this up - ie mobile). Despite their similarities there is a HUGE challenge to meet the user's expectations when you are not fixed in one model.
There are a few different points of view I see in the industry - one is that people will get better at the touch language and our interfaces will evolve to support complex touch interactions. And some who think that we have achieved the optimal method for efficient interaction and and the optimal method for mobile quick interaction and that the two worlds should be separate. And still others who believe the best interaction method is still out there (gestures, voice, pen, natural language, ai etc.)
The view of windows 10 is that these two interaction paradigms or mental models support two semi formed methods of computing that are continually converging. The attempt is to create a UI that is adaptive to the users goal and should support: WIMP, FIM or a crossover between the two. The closest analogy is definitely responsive web design.
There are huge drawbacks to this such as complexity, and current user expectations. But also huge benefits such as computing that is less tied to a device or input method, and lets you choose the best tool for your task at hand.
I think it will be very interesting to see how the market reacts to this convergence, and if microsoft is able to communicate real user benefits from this type of experience. So far, it seems like simplicity is key.
As a tech geek and designer, I'm excited for the next 5-10 years of interaction design. I think a lot will change with new tech (AR/VR + natural language + ai) and the continual evolution of our existing UIs. I don't expect many major shift within already established interaction paradigms, but shifts that will influence them from outside technology.
edit: there are so many facets to this that I couldn't capture them all - but I wanted to comment on the card metaphor. To me it's a symptom of designers and devs looking for a metaphor that is able to cross between devices, screens and surfaces and escape the boundaries of the WIMP methodology. The fact that these types of metaphors are arising naturally in the world of software is a signal that the zerox/parc metaphor approach is still the best for mass market adoption. And seem to be the most well adopted to the environment we have created.