There's a new content discovery site going into private beta today: Thoof. The concept is that it knows who you are, where you are, and what you like, and picks online sites and stories for you based on that information.
Thoof users still have to submit stories to the site, and right now the only way to do that is to paste a URL into the system. That will slow down submissions, compared to having a bookmarklet or a similar quick-add feature. But I like the idea of an aggregation site that treats me as an individual, not just part of a crowd of users.
CEO Ian Clarke says Thoof knows what you're going to like, "from the first moment you visit the site." There's a lot of information feeding into the system. For example, it takes all the information your browser passes to it, such as your platform and IP (which it uses to take a stab at your physical location), and correlates that with items that other similar people have clicked on. Of course, it also records what you click on, but it doesn't yet develop a profile based on your Web history outside of Thoof or from other clues you leave on your system.
Thoof also keeps its own ontology of tags (all items must be tagged when they are submitted) so it knows what tags are related to each other. If you click on a story, Thoof will, in the future, recommend other items that are related to it by tag. Over time, the system can also set up negative associations. For example, if you like Formula 1 racing it might initially give you Nascar news as well, but should learn quickly that Formula 1 and Nascar is actually a strong negative correlation, not a positive one.
Thoof will depend on its user base to stock it with good links, and as I said, that's a weakness. Other news aggregation services, like Google News, are automatically stocked from RSS feeds and Web crawling. Clarke said turning the "autopopulate" feature on is a simple thing for Thoof to do, and he will do so if necessary. I think it will be.
I also have an issue with the basic concept of Thoof. While it's great to get personalized recommendations, sometimes you want to know what your friends are reading, so you can join in the conversation. That's what Digg provides so well--a collection of stories that you know other people think are interesting. It's a shared experience. But Thoof is solitary. You read stories that the machine thinks you'll like, and you have no way of knowing if other people like them, too.
Ideally we'd have one system that gives us both algorithmically-generated recommendations and group recommendations. I'm still waiting for that site to launch (if it has already, please let me know).
I have 25 invitations to Thoof that I'm giving out. They're gone, sorry.