One of my favorite companies from the January, 2008 Demo conference was Delver, a search engine that takes into consideration who your friends are and what they've said and bookmarked in its results. I liked the idea in January (see ) and I like it still--I've brought it up in several posts since then.
I finally got a chance to try it out, as you can today, now that the site has gone into open alpha testing. At the current time, it's cooler in theory than in practice, but there is a ton of potential here.
Once you tell Delver who you are, it builds your social graph by itself. It correlates your identities across sites like MySpace, LinkedIn, Flickr, Facebook, Digg, and more, and creates a list of who your friends are. It also layers in a list of your friends of friends. Then, when you search for something, it gives precedence to content and links from your friends and their friends. So if you're looking for an Italian restaurant recommendation in New York, you'll get results from people you know, or people your friends vouch for. Delver CEO Liad Agmon clearly believes that online content from people you know is more valuable than generic Google results. He says, "The Web is no longer just a collection of documents. It's made up of microcontributions."
You can also help Delver build your social search graph by feeding it sites and blogs where you hang out and have friends. Or to really supercharge it, feed it your FriendFeed page, where the "fusion" of multiple online personalities is already coded in. But you don't have to do this, and that's one of Delver's very cool features: It discovers your network on its own.
Delver can only extract results from people who post content or link to Web pages. If your friends are quiet online, you won't get much in the way of relevant personal responses. But you can, if you like, define anyone as your "search buddy," and then when you search, the results they would get become yours. You can even set up multiple search buddies to blend together the results from several people who have good online networks.
In using Delver, I liked how it told me the relationship of the person whose results it gave me. Some results came from my direct contacts, some from friends of friends, and some were selected because they were from people who went to the same school I did or worked in the same company.
However, in the alpha, I often got random (non-friend) results ahead of results from my social circle. Also, Delver doesn't index Twitter, and won't directly do so, according to Agmon. All those tiny posts would clog the Delver engine. Instead, eventually, Delver will clump Twitter posts into groups and index those intermediate pages. (Which doesn't explain how Summize manages to index Twitter.)
Coming later, possibly at the TechCrunch 50 event, will be a widget for bloggers: a "grey-label" search solution that gives blog readers an opportunity to get search results filtered by writers' social networks. Agmon says this will let bloggers "become prisms to the world," for their readers. We've seen custom search engines before (Eurekster, Rollyo), but this does sound like a nice add-on product for Delver.
Delver is well-funded, which is important since search is an expensive problem to solve. The company will make money the old-fashioned way: From search advertising. Agmon has not yet revealed which advertising network his company will use.
All Webware stories mentioning Delver