AI recommendations
An AI movie recommender that shows its work.
Every recommendation engine claims to know your taste. Almost none will tell you why it picked something, which means you cannot tell a good guess from a lucky one — and you cannot correct it when it is wrong. Kinolog cites its evidence: every pick names films from your own diary, so the claim is checkable against what you actually rated.
Reasons you can check
“Squarely in your wheelhouse. The same science-fiction pull as Inception” is a different kind of sentence from “recommended for you”. It names a film you rated, so you can agree or disagree with it. When the engine is reaching outside your usual taste it says so and labels the pick a stretch, rather than quietly mixing it in and hoping.
It also refuses to overclaim. A diary of six films does not support confident statements about anybody’s taste, and Kinolog will tell you it is still learning instead of manufacturing a personality for you. How AI movie recommendations actually work — and when to distrust them goes through the failure modes in detail.
Built on your diary, not on what is popular
The engine reads what you logged: ratings, the vibe tags you chose, whether you would rewatch, and the notes you wrote in your own words. Notes matter more than they look — “too long but the score carried it” tells a recommender something no star rating can. Films you have already seen, saved or ruled out are excluded in code before anything is shown, so the same title cannot come back around.
Picks for tonight, not a ranked list of everything
Today’s set is five films, composed once and steady all day so a decision does not move while you are making it. It spans intentions deliberately — something comfortable, something from your watchlist, one deliberate stretch — and no two picks lead with the same genre, because five good picks that are all the same thing make a bad evening. Filters for time available and mood re-rank instantly without regenerating anything.
Permission first, and a real answer if you say no
Nothing from your diary reaches a third-party model until you agree, the provider is named rather than described as “our partners”, and declining is a real button rather than a dismissal. Say no and the diary, search, stats, Wrapped and your daily picks all keep working on a local engine — you lose the AI-composed picks and nothing else.