http://www.cincomsmalltalk.com/blog/blogView?showComments=true&entry=3232986164

especially the comments thread. This is nothing new - except, apparently, to Gosling

One thing about Gosling’s particular take that was new to me [no doubt its laughably old within the community on the above-linked forum (c:] is the amount of automation. It sounds like he wants the machine looking for bad smells in the AST and proposing/performing the necessary refactorings to remove them. As he puts it, “trying to couple analysis with action.” Since a lot of refactoring starts from an existing mess, this sounds like useful assistance.

I kind of like the idea that an analyser in my development environment might look at the existing tree, and notice ways to transform it to improve some metric related to simplicity or comprehensibility. It seems that the “best” architecture often evolves in fundamental ways within real-world codebases that are chasing a moving target. To the extent that a good architecture is one where classes map nicely to concepts, then such an analyser might even help me to usefully reconceptualize the problem that my code currently solves (and the way that it solves it.)