Chris Pirillo of Lockergnome fame just launched gada.be, a metasearch service (beta, of course), based on RSS feeds and the results are output in OPML (Outline Processor Markup Language). It's simple and quick, which is great for searching on mobile devices. Currently Gada.be searches about 150 feed sources. Google, for example, doesn't offer it search results as feeds, so the universe is limited currently, but that's likely to change soon. Gada.be is a good example of innovations (in this case germinated primarily by Dave Winer) that help define the notion of Web 2.0.
At the other end of the metasearch scale, Inform.com, munches on content from about 1000 news and 100 blog sources, normalizes it and extracts concepts such as topics, people, places, organizations, industries and product. Inform also has a method for scoring the relevance of articles based on the concepts in an article. According to Neal Goldman, CEO of Inform Technologies, the goal of the metasearch site is to "Tivo the news for people," Goldman told me.
When you search on a term, you get relevant results and on the right side of the page the related concepts. You can drill down at the intersection of the search term and a concept, such as a related organization (Microsoft and Google), or fork down the path of the organization (Google). The concept modeling and other techniques that Inform.com applies to news and blog content aren't new. Topix.net similarly aggregates news, but its 'related pages' feature isn't as sophisticated as Inform's related concepts.
Currently Inform doesn't produce RSS feeds, but Goldman said they would in the next month or two. Inform is definitely in beta and doesn't support Firefox as well as Explorer. The company is planning to generate revenue via targeted ads on its search results pages (not the content pages), and to offer subscriptions for services for personalized features, Goldman said. Later, he said that the company would add video sources.