No doubt, .NET and managed languages are helping a lot to TRY to implement things faster. No doubt, native code for sure IS faster at runtime, doing interactive things in UI. 

Can you imagine that WPF, in fact, is quite quickly built managed prototype of full-featured XAML implementation on top of Win32/GDI+/etc..., that Silverlight (his "agcore" part) is native rewrite of this WPF prototype (slowly filling all his capabilities...), scripted by lighweight/portable NETCF/CoreCLR (also in fact rewrite of bigger full .NET prototype, originally bound to Win32) and NOW its time to bind this declarative UI also as first-class thing ("Jupiter") into new windows, learning also from "Silverlight embedded", already capable of C++ and from light WinCE core?

Now, we have well-behaved hw accelerated virtualization, hw accelerated graphics etc.

I think that Win8 will have new lightweight kernel with tightly coupled CLR (SLR? even more native rewrite of core things prototyped by CLR?), tightly coupled hw virtualization support (why not to run legacy apps seamlessly using VM as did in "WinXP Mode"?), tightly coupled with even more native rewrite of WPF/Silverlight (Jupiter?), for sure powered by declarative XAML forever. And all this powered ALSO by new WinC++ (and even JavaScript) bindings to .NET-like APIs/syntax.

XAML declares WHAT you want, but never HOW to do it. This declarative paradigm will continuously hit more and more parts of applications in future. What about .NET also as big prototyping tool in fact?

Nobody is smart enough to predict and design silver bullets, all software development is continuous try/fail/enhance loop, not to mention other companies and all social(istic) HTML# crap without any innovative ideas how to simplify development and move to something REALLY better.

BTW, even "way can be goal" and I like MS way.Smiley)