I rebuilt the "similar movies/TV shows" algorithm on TasteFinder

Name: I rebuilt the "similar movies/TV shows" algorithm on TasteFinder
Availability: InStock
Author: tastefinder_io

by tastefinder_io·Mar 12, 2026·3 points·0 comments

Visit Project View on HN

AI Analysis

●MidCrowd Pleaser

Multi-signal matching beats genre tags, but Letterboxd already dominates this space.

Strengths

•Matches tone, pacing, and narrative structure beyond simple genre overlap.
•Tinder-style swipe interface makes rating frictionless and addictive.

Weaknesses

•Recommendation algorithms are hard to verify without large user datasets.
•Movie discovery is saturated with Letterboxd, IMDb, and streaming platforms.

Post Description

Hi HN,

I posted TasteFinder here before - it's a movie/TV discovery tool where you rate titles with Tinder-style swipes ( / / ) to train recommendations.

One thing people kept asking for was better “similar to this” results. So I recently rewrote that part of the system.

The new similarity algorithm tries to match titles across multiple signals instead of just genre overlap. Things like: - tone / vibe - pacing - narrative structure - themes - audience preference patterns

The goal is simple: when you search for movies or shows similar to something you love, the results should actually feel close, not just share a genre tag.

Example use case: You finish a show you liked and want something with a similar atmosphere — not just “another sci-fi”.

If you're curious how well it works: https://tastefinder.io

I’d really appreciate feedback on the similar titles results specifically. If you find cases where they’re great or completely wrong, that’s especially useful.