Trapit, a personalized tool for discovering Web articles, opens to the public today. Trapit crawls roughly 100,000 sites, adding more sources every week, to provide users with the most relevant content from deep within the Web, not just the popular or SEO-spammy results. It's built on the same AI technology as Apple's Siri, which means it learns what interests you and gives you better suggestions over time.
You enter a search term for whatever you want, which you can save as a "trap" that will automatically refresh with new content as it's published around the Web. Every time you log in, you'll see new stuff to read, and the suggestions get more personal every day. The Web app launches today at trap.it, but Trapit was developed as a platform, so this is only the first stage. "We expect to power sites and services across the Web," CEO and co-founder Gary Griffiths says.
When you search on Trapit, the first batch of stories might be pretty good, depending on your query. The interface prompts you to give five stories the thumbs-up or thumbs-down until it's finished personalizing. This isn't an up-vote or down-vote for popularity; it's just whether the article is what you're looking for or not. This is how the AI engine learns what you like and personalizes results for you. After that's done, the results are fine-tuned to your tastes, and the same trap on someone else's profile might look completely different.
Don't think of Trapit as a search tool. You can save traps to your profile, and as the engine finds new stories it thinks will interest you, it delivers them to your traps and gives you a notification. Trapit makes for a great homepage; every time you open your browser, you'll see new stories listed in your activity feed, which you can read now or save to your reading list for later.
While the trendy discovery engines these days are social, trawling your Facebook and Twitter connections and using those to approximate your interests, Trapit goes the other way. It uses only your query, your votes and its machine intelligence. "There's no concept of crowd-sourcing on here," Griffiths says. Trapit shows you featured traps by other users, which you can add as your own, but as soon as you do, they start personalizing for you specifically.
Trapit reminds me of Thoora, an app with a similar mission, but they work rather differently. Thoora's algorithms use certain signals, including popularity but also using smarter semantic data, to pull in content from millions of sources. Trapit scours fewer sources, but it uses different underlying technology with a grasp of natural language.
