I agree that GWT is not friendly to UI people who are used to writing their own markup. But I would argue that a good UX person should be concerned with how the user interacts with the application (not necessarily by writing HTML and CSS by hand, but by sketching out the design on paper or Illustrator), and a framework like GWT often makes it simple to build complex UIs that would be difficult/labor-intensive to create and maintain with a traditional web dev stack. A decent developer should be capable of taking mockups from a designer and building out the rounded corners and other pretty bits himself in CSS.
Everyone has their preferred toolsets and frameworks and of course they should use what they feel is most effective. Case in point: my shop uses very little GWT - my coworkers prefer different tools (and often so do I; GWT is overkill IMHO unless you're building something big).
But this is all not really relevant to my point:
I was talking more about how the linked-to blog post was kind of web-development-101 stuff that everyone already knows and somehow, sadly, still received many upvotes/comments. I only brought GWT into the mix because some guy wrote a blog post about listening for URL hash changes and is presenting it as the future of web development, while (using GWT as an example) people with PhDs have written an optimizing Java to Javascript compiler and engineered very good solutions to difficult client-side web development problems that completely trivialize something as basic as hash change history tokens.