Method · machinery · fine print
The critic is made of math and bad manners.
Verdict is a tiny, opinionated movie-taste test. It is not a recommendation engine, a psychological assessment, or a substitute for arguing in the group chat.
How the verdict is made
The machinery behind the insult.
Ten reactions, five axes
Love, hate, and skip signals are translated across craft, strangeness, prestige, emotion, and pace. Skips help the deck adapt; opinions shape the score.
A deterministic verdict
Your taste vector is matched to the nearest of eight named archetypes. The same films and reactions produce the same diagnosis—an LLM does not choose the result.
AI supplies the raised eyebrow
Groq-hosted GPT-OSS 20B may rewrite the verdict and prediction copy in the critic’s voice. If it is unavailable, the scored result still works with editorial fallback copy.
Privacy, without the monologue
No account. No taste dossier.
Your card reactions are used to calculate the current verdict. CriticWise does not write them to an account or user-profile database.
When AI copy is available, the loved and hated movie titles—not an account identity—are sent to Groq to rewrite the prose. The score and archetype are calculated before that request.
Cloudflare Web Analytics measures aggregate site traffic and performance. It is not used to build personal movie profiles.
Data & credits
The names in the small print.
Movie metadata, posters, backdrops, trending titles, ratings, and trailer references are supplied through The Movie Database API.
This product uses the TMDB API but is not endorsed or certified by TMDB.
Verdict is designed and produced by CriticWise. Taste-model scoring and fallback copy are part of the application; optional prose enhancement runs on Groq-hosted GPT-OSS 20B.