Last week, I got a spam comment that included a bunch of urls. I told pLog that this comment was spam, and it trained its Bayesian engine on it.
Today I got a couple more attempts to post spam comments, and they were all blocked.
I am pretty impressed with pLog's implementation of it's Bayesian anti-spam filter. It actually uses the posts that the author makes for training as not-spam. This is great because normally comments are related to the content of a blog and, at least on this blog, there are a lot more posts than comments.