The grand experiment has begun.

The problem: I've been on a project developing a very deep application for four years now, and lately it's become so complex and intertwined that things are starting to break. Developers often say "The application is smarter than I am." Throughout all of this I started to notice that many of these bugs could be found if we had ... automated GUI tests.

The question: Why is automated GUI testing a bad idea? I've read some articles that discuss why it's bad, but can some (most?) of the bad problems be overcome?

The cause: Create a GUI testing application that is so simple and powerful that you can't resist using it.

The name: Not much imagination here, I'm just going to call it eXtreme GUI Tester, or XGT for short. Following some ideas from eXtreme Programming, I'm going to turn all the dials to "10" and see what happens. Maybe it blows up, maybe it works.

The code: Well, that's coming soon. I can say that I'm in deep development, it's being used on live applications, and it's going well. As for the external release all I can say right now is "please stand by".