The Pipes interface is stunning. I want to use it just because it looks nice. Once you get your head round actually using the interface (why can't I double-click to add modules, as well as being able to drag them?), creating new Pipes is dead easy. Just as a trial run I created a very basic All About Milton Keynes pipe that brings in items tagged milton-keynes on del.icio.us, recent changes from the Open Guide to Milton Keynes, and things mentioning the words "Milton Keynes" from Revyu. It actually not very useful right now (the challenge is now in the novel combinations, not the technical glue), as it mixes different types of items (links to web resources, city guide entries, and reviews) without distinguishing one type from the other. Maybe that's down to me choosing to aggregate all items and sort them alphabetically, but actually the real story is that this is just the limitation of RSS compared to RDF that uses a range of vocabularies.
Annoyingly, as I haven't finished the tag based RSS feeds on Revyu (unlike those that are People-based, like mine) I had to resort to using the generic latest reviews feed and then filtering it by keywords. Not right for a Semantic Web application. Really I just need to put aside an hour or so to finish the tag-based feeds, but I kind of object to having to do that when Revyu already has much richer information already available in RDF/XML. So, there are two ways forward. Do the feeds, and deal with it, or wait until Pipes begins to speak more RDF than just RSS. This is probably a way off, but would be awesome. Being able to create FOAFmap (currently offline) style Semantic Web mashups by dragging much richer data than RSS can provide would do wonders for bridging the gap between the Web2.0 Wow! factor and the web-scale data integration capabilities of the Semantic Web.