Jybe is a social recommendation app for the iPhone that gets smarter the more you use it, somewhat like Pandora or Netflix.

Initially, it focuses on three categories – restaurants, movies and books, and there’s a good reason for that.

Unlike much of the information available online, most of the digital data about restaurants, movies and books is structured in such a way that it is relatively straightforward for a technology like Jybe's to organize and display it.

Jybe’s co-CEOs, Arnab Bhattachariee and Tim Converse, both spent more than a decade in the search business, at companies including Inktomi, Excite, Yahoo, and Microsoft.

“In search, if you do a good job of entering key words, the search engine will nail it,” says  Bhattachariee. “That’s the explicit mode of search. But there’s an implicit mode as well. There are lots of things of value that you can find with serendipitous search.”

In order to help you discover those things, Jybe first needs to know a little about your tastes.

“We ask a few minimal upfront questions about books, movies and restaurants, and we make it engaging and fun,” says Converse. “On the first screen we show you movie categories and you click once, twice, or three times to indicate if you like it, don’t like it, or are neutral.

“On the second screen, we show you actual movie titles in a category you like, say SciFi. We’ve found that people tend to zero in on what they like and don’t like and then they are done. That’s good from our perspective – we want to know your strong likes and dislikes.”

With this kind of machine learning, users need to train the algorithm – thus the upfront questions.

“This on-boarding data is substantially more useful than the average person’s Facebook profile, which does not contain so much useful data,” says Bhattachariee.

“That’s because our questions yield data that is comparable across our site,” explains Converse. “Then we mix in discovery with the specific results. There’s a term in machine learning -- explore vs. exploit – which means we provide a mix of what is up your alley with what’s up the next alley over.”

With restaurants, the two-screen on-boarding questions start with categories of cuisines, then with appetizing photos to specific dishes.

“Your stomach reacts to the screen,” says Bhattachariee. “Especially when you are hungry.”

Jybe uses this data not only to recommend specific restaurants, but the dishes you might like at those establishments.

One surprising feature with Jybe is its “cross-domain recommendations.”

“It turns out that if you like a certain food, other people who like it will also like similar movies to the ones you like,” says Converse. “People’s tastes tend to cluster across the categories.”

Once you do find something you like with Jybe, the service helps you take action, by making a reservation or a purchase.

As for its social layer, based on your profile and your social graph, Jybe offers you the option to follow people who tastes are similar to yours. There’s a cool taste dial that reveals how similar to someone else you actually are, at least in these three categories of items.

The company has a team of six and is headquartered in downtown Redwood City. Although only in iOS right now, Web and android apps are coming.