Nice. What algorithm is used for recommendation?
It gets embeddings from open AI and compares them using cosine similarity. Stories are weighted by similarity, dissimilarity and HN score.
How you compare a liked set to single embedding? By best match? Or is there some fancier way?