I used MovieLens, but got frustrated by how generic it felt. Most recommenders rely on collaborative filtering, which naturally suffers from popularity bias—popular titles dominate the math, so you mostly get fed well-known films.
To fix this, I built a movie recommender using Physarum (slime mold) vascular conductance. Instead of standard matrix similarity, it models taste as fluid flow across niche curator clusters and dampens high-traffic blockbusters so obscure films can actually surface.
