It has a way to go, but already is fairly mature, you can see more here:
http://www.microsoft.com/expression/Default.aspx

Potentially, the focus on code could constrain the final results and hinder some creativity. The ideal approach I’ve found is to iterate, and work between paper and code. So use paper sketches to explore ideas, and iterate your code with the best ideas.

Exactly – that’s the issue with Omnigraffle, Fireworks and other tools – you end up throwing out the prototype. But there are a stack of online tools around that speed up the process and are handy for quick client communication, and can be used in place of Omnigraffle, Fireworks and others. While the results from online tools are also thrown out, they’re a substitute for at least part of the typical client conversation. Of these, http://iplotz.com/ seems to be one of the more efficient and practical services.

It’s ok to throw out a prototype, as long as it’s made to be thrown out and has been super-fast to produce. Quick and sketchy – on paper or in tool like iPlotz – usually fit the bill well. Put something together quickly, chat to the client, iterate until it’s right. Then code. Then iterate from there, but the purpose of the ‘sketchy’ prototype is to avoid completely recoding a layout! If designers are involved from the beginning, you can wireframe with them and you can agree on the layout early. Then the art direction can be applied to your coded prototype and avoid too much re-work.

I’m taking a look at Omnigraffle and Fireworks, but can see myself spending too much time architecting wireframes that will as you said, be thrown out. Yet, I can also see myself having to recode the layout once the designer changes things up a bit.