On Persai

I’m not one of those people who thinks any new startup in search is bound to be beaten by Google. It can’t be done by playing Google at their own game but it can be done by using a different approach. Search startup Persai falls into that category. What struck me was this description from co-founder Ted Dziuba:

We want to build machine programs that can learn things from information that’s out there on the web. In the first application we’ll come out with, you tell us things that you’re interested in, and we’ll continuously go out and find stuff on the internet that’s related to that. There’s a positive feedback loop where you tell us what you like and don’t, so the machine gets progressively better in learning what you like.

I like the basic idea but there is one problem: I don’t really want to tell yet another site what I’m interested in. I’ve told delicious and Digg what links I like. I’ve told facebook what films and books I like. Why should I tell Persai all over again? If Persai could work out a way to yank all this stuff from the other sites automatically and knit it altogether then serve relevant results that would be a fantastic tool, unfortunately:

Recommendations are based entirely on content, other users’ feedback has no bearing on what Persai recommends to you.

This is a mistake. I’ve said before the key to next generation search is analysing the links between users as well as Google analyses the links between pages. Persai seems to have it half right in that I can tell it what I like but I think they need to rethink their approach on other users’ contributions.

There are a few startups looking to revolutionise search for a second time. I met the founders of another new search engine Piins a few weeks ago, they are still in private beta so I can’t go into specifics but they also have a very interesting concept which has a good chance of success. Both Persai and Piins come into the “companies I would definitely invest in” category, if I were a VC.