Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

It's impractical even for tinkerers. YouTube claims to get over 20 million videos uploaded daily and it has well over 10 billion stored videos in its corpus. The metadata alone is tens of terabytes. The usual introduction-to-recommendations approaches out there are going to completely fall over on an item set of this size, even if you have disk space to spare.


The server can deliver a sparsely randomly sampled RSS feed of embedding vectors and metadata.

Fetch media after ranking on-device.




Consider applying for YC's Winter 2026 batch! Applications are open till Nov 10

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact