If RSS is really "TiVo for the internet", when does it start making recommendations? TiVo is basically an aggregator, just like NetNewsWire. It keeps track of the shows you "subscribe" to, checks to see when new "items" are available, and downloads them for you to look at when you're ready.
One thing TiVo does that NetNewsWire doesn't is recommend shows you might like that are not currently amongst your subscriptions. No aggregaror does this, as far as I know. Although these recommendations don't always work perfectly and you may get stuck with TiVo thinking that you're gay (registration required), they seem like a pretty good idea.
Our aggregators know a lot about what we like. Mine knows that I read a lot of posts about music distribution, intellectual property, arts administration, and animation, and a lot of other posts that have podcasts from public radio shows attached to them. If NetNewsWire maintained a database of RSS feed providers (or collaborated with someone who did) it could tell me when someone started a new podcast about the administration of animation non-profits, say, or even if a single post appeared on some obscure blog about a new Creative Commons-licensed mp3-sharing site. If NNW, or one of the other news readers, didn't want to do it, it wouldn't be hard to create an online "RSS Registry" where people could submit their feeds with keyword descriptions and their individual posts with Technorati Tags via ping. Then a user could go to the site, enter the list of feeds they subscribe to and receive in return a new RSS feed that notifies them of new posts and new blogs that the Registry's recommendation engine thinks they might like. The big disadvantage of this system (that you would have to constantly return to the site to keep it updated on the feeds your subscribed to, weakening the feedback mechanism which would make it truly powerful) might be mitigated by the use of aggregator plugins that would keep your Registry account up to date. The plugin could also do things to refine your recommendations like keep track of the posts that had links you clicked. A big advantage of implementing the recommendations through a website rather than a service provided by an aggregator might be that the site could take advantage of the social network of its users to give better recommendations (i.e., "people who subscribed to this feed also read. . .") in addition to taxonomized (or, maybe, folksonomized) labels.
Some people might be concerned by the idea of their RSS reader starting to talk back, sending information about their interests off to some sketchy outside entity, just at a moment when so much brainpower is going into trying to figure out how to make money from syndication. But I think that there's something really powerful to be gained in the tradeoff: closing the feedback loop, getting more and more useful information with less work. And privacy concerns are just more of a reason to implement something like this yourself, so that it will be free, fair, and open rather than owned by one of the aggregators or someone even less trustworthy. Because with all the work, thought, and investment from people a lot smarter and more knowledgable than me going on in this space, someone else is going to think of this before too long and who knows where they'll take it.