Niall Kennedy spent some of his holiday time reverse engineering the API behind Google's Webfeed Reader. Apparently, it's a fairly straightforward REST based API that supports easily retrieving and managing subscription blogrolls and unread items. You can also leverage the feed and item tagging that the reader supports.
The big bonus is that Kennedy got the Google Reader team to confirm that they intend to open up the API for independent developers. Timeframe looks like Q1 2006 for something official. Lot's of potential here for interesting new aggregator experiments, especially if blog search was eventually made available, although I realize that last bit is a tricky design and implementation issue.