Trapit is an interesting little service that acts as a hybrid between a search engine and a recommendation app, with a number of social components that go on top. Today the startup is introducing an iPad application to give users access to their content directly via a mobile client.  It’s one of the biggest updates from Trapit since its beta launch last November, addressing a key market for growth and one that I’d noted as important for Trapit’s ongoing development in an initial review of the web app.

Trapit aggregates data from about 120k handpicked web sources and applies an AI built on the same code as Siri to customize this stream based on users’ individual preference. It tracks things like browsing history and other engagement metrics, while also making it possible to add articles to a reading list in order to take full advantage of this underlying functionality, essentially hoping to replace Safari as your default browser when perusing Trapit content.

The platform is integrated with Facebook and Twitter and lets you share your favorite tidbits from around the web. The iPad app features a lot of the capabilities found on the web, as well as a number of other things. Gesture shortcuts are supported, and there’s also a ‘New Discoveries’ feed that displays all the user’s Traps and social networking integration.  You can add Traps to your reading list for later viewing, easily accessible on your home page.  The browsing experience in general is very fluid and easy to navigate, especially compared to similar apps that feature news feeds and web content, often cumbersome to switch categories or flip through articles.

“Personalization has become nothing more than a buzzword,” said Hank Nothhaft, chief product officer and co-founder at Trapit. “Personalization should be about you, the individual, and your unique tastes and interests, not about what your friends are buzzing about on Facebook and Twitter. So we’ve built Trapit to provide an experience that is unique and carefully tailored to each user. That goal, enabled by the underlying technology, and our distinctive approach to content sourcing is what makes Trapit so compelling to our users.”

It’s evident that Trapit took great care in the design and usability of the app, making it very similar in function and UI to the web app while also taking iPad-specific capabilities into consideration.  One notable feature is Trapit’s expanded media format, which loads full web content for a selected link directly in the app.  This keeps you from having to jump from Trapit’s app to Safari, but works far better than your typical “skinned” browser page.  In fact, you’d never know you weren’t using the iPad browser, save for the fact that certain functions like your bookmarking capabilities aren’t yet integrated into Trapit’s webpage viewer.

Mobile just a piece of the big picture

The mobile move is a smart one for Trapit, but it’s only one development amongst many the startup has in the pipeline.  Trapit’s been taking a steady approach to its feature roll outs, updates and partnerships, focusing heavily on the core technology behind its algorithms and AI.

It’s this focus that prompted me to ask Nothhaft how Trapit maintains data integrity with  his beloved AI, considering its underlying technology is, after all, still designed by humans.  The matter of data integrity is an important consideration for Trapit and its competitors, Siri and every other service that essentially programs an AI to learn on its own.

“We don’t have a set database of content, Nothhaft explains, “we have the same flexibility in Trapit as a search engine.”  It’s at this point Nothhaft pulls up a few rival apps and keys in a search, showing a string of categories to follow instead of being able to create a category around his query.  “It takes a human to determine that vocabulary and it’s not supported in their systems.”

“Even though we have humans behind the sourcing and the algorithm, each search on Trapit creates its own ontology.  [Users] are creating the vocabulary with the keyword and URL, and we build it from there.  And it works,” Nothhaft says.  “These others don’t learn.  They work with fixed taxonomy and ontology…there’s content that might be relevant but it’s inherently limited.  We architected from the ground up for total personalization.”

Contributors: Maria Deutscher