Kevin Lang and Vivek Tawde of Yahoo! Research constructed an RSS reader that clusters items based upon a fixed set of topics. The aggregator runs as part of the Yahoo! Widget Engine so it's essentially a desktop RSS reader. Also, from my read of the project description, the selected sources are rigged or at least initialized to a providential set. However the task is still quite tricky since the cluster sizes apparently aren't fixed and have to be learned from a given pool of content.
Another piece of The Daily Me drifts into view.