Mouthful of a title, I know.
During my reading-USENET-via-nn days, I envisaged a news reader that would learn from what I selected and what I didn't select to read and would sort the articles according to how likely it thought I would want to read them.
I didn't know about Bayesian Classification at the time. Now that I do, it seems the perfect technique to use.
I wonder if a similar technique would be useful in prioritizing the reading of blog entries. Admittedly, the signal-to-noise ratio on the blogs I read is considerably higher than USENET but the quantity of blogs I now read makes it potentially useful.