You can implement what you want with simple keyword & url filtering.
What I'm talking about here is uncovering "latent" communities, if you will. As in, make a giant matrix with the users being the columns and the posts being the rows and then use the eigenvectors to make recommendations (see SVD: http://en.wikipedia.org/wiki/Singular_value_decomposition)
The benefit of this approach is that I no longer have to be conscious of the topics I am filtering in or out. Even keyword based filtering is, again, a coarse estimation of relevance. I may be very interested in clojure, but I'm certainly not interested in every article that contains 'clojure' in the title.
An SVD (or similar) approach would filter my interests loosely on the co-occurrence of votes. That is, a vote from someone with whom I have high overlap is worth more to me than a vote from someone with whom I have never voted the same direction on the same post.
In any case, co-voting data is not scrape-able from the public HN site, so I think using keywords and urls is really the only realistic filtering option at this point.
Here's a screenshot of one of mine http://imgur.com/NLOkM.png
I wish more sites had that kind of filtering.
Think of it like email spam. You can setup manual filters to filter out email spam, but that is a constant and never ending stream of work for you. A simple bayesian filter like pg has described will require far less work and give far better results.
In this case, a machine learning approach is even better because it can bring up stories that a user will be very interested in even though the story would never make it to the current homepage.