ANSWERS.

An anime recommendation engine that learns from how you swipe. You rate shows by swiping left or right - or superlike the ones that actually hit - and it builds a taste profile from that behavior. The longer you use it, the more accurate it gets.

There's also a directory of 18,000+ anime you can browse by genre or tag, and individual pages for every show with ratings, signals, and contextual recommendations.

Yes. Free to use, no account required to start. An account lets you save shows and keep your taste profile across devices - that's it.

No. Your taste profile is stored locally and works without signing in. An account syncs your profile and saved shows across devices so you don't lose your data.

18,310+ and growing. Every show in the catalogue has its own page. The engine has signal data on the majority of them - the more data a show has, the more precisely it can be recommended.

You can search in the directory. But the recommendation engine doesn't use search because search assumes you already know what you want. OTAKULT is built for the question you can't answer yet: what should I watch next?

Swiping gives the engine more signal than a search query ever could. Every choice is data. Data becomes taste. Taste becomes recommendations that actually land.

Every anime in the catalogue has a vector of signals - values derived from how real viewers described watching it: emotional intensity, binge pull, visual craft, darkness, pacing, divisiveness, comfort factor, and others.

As you swipe, those signals combine into your personal taste profile. The engine continuously compares your profile against the full catalogue using cosine similarity, weighted toward the dimensions that matter most to you. What surfaces first is what the math says fits best.

A vector of signal weights that represents who you are as a viewer. Not your genre preferences - your actual experience preferences. High emotional intensity. Low comfort factor. High binge pull, medium pacing. That kind of thing.

It builds passively as you swipe. You can see it in your profile.

A superlike carries more weight than a regular like. If you superlike a show, the engine treats its signals as stronger anchors for your profile - it learns more aggressively from those choices than from a standard swipe right.

Use it for the shows that genuinely hit different. It's not just a stronger like; it's more information about exactly who you are as a viewer.

Keep swiping. The engine needs volume to calibrate. The first 20 swipes are rough. By 50 it has something real to work with. The profile improves with every choice you make - likes, dislikes, and superlikes all contribute.

If something is genuinely off, the dislike signal is strong. Skip things you're neutral about; swipe left on things you actively don't want. That distinction matters to the engine.

The engine at /discover is the active recommendation experience - swipe-based, adaptive, personalized. The directory at /browse is the full catalogue with genre filters, tag filters, search, and sorting. They're separate surfaces built for different modes of looking.

Every show in the directory links to the engine so you can use something you find to seed what comes next.

Those are the show's strongest signals, translated into plain language. They reflect what the engine actually knows about the show - not just what genre it is, but what it's actually like to watch it.

If you check them and click "Taste Profile →", the checked signals get used to pre-seed your taste profile in the engine. It's a shortcut to telling the engine exactly what you liked about that show.