Persai: dull but effective
As I suppose quite a few people did, I signed up for the beta of Persai as a reader of the now defunct uncov. I imagine that gives them a fairly direct batch of beta testers, all over-fond of the word 'fail'. I've used it solely from an end-user point of view: I haven't done any testing to try to figure out the technology yet.
Persai itself, is almost the diametric opposite of uncov in character: quiet and unassuming, and distinctly unchallenging.
At heart, it's an RSS aggregator, with the slightly more interesting twist that it shows you feeds based keywords rather than picking URLs, and the hook that it is going to learn what you like based on some kind of linguistic analysis of the words you choose, and predicting which new articles you're mostly likely to read. You get a 'reject' button if the matching didn't work out.
It seems to do this rather well. I picked a fairly broad range of keywords and it genuinely seems to return articles in a much more effective way. For example, one set of keywords I entered was related to philosophy, and frankly, not many new articles appear - but this is a good thing, since frankly, there isn't exactly a wealth of philosophical discussion in the news - although there are a large volume of articles using the word philosophy daily, and that's the type of unhelpful response I'd get from a pure keyword rss aggregator like Google News.
The interface itself is extremely plain, and there's very little that could be described as attention-grabbing about the whole process. There's a danger that I might forget that it even exists. But I get the feeling that Persai is trying to do one thing very well, and is doing a very reasonable job of that thus far.