I "wrote" (vibe-coded) a non-chronological RSS reader. I crawled the blogrolls of some blogs I liked, fetched all the posts (with full text) and ran a clustering algorithm on them. LLMs classify the content to filter out meta/political/personal posts. And a LLM gives names to the clusters by sampling the contents.

It's an amazing source of long things to read. There is so much stuff worth reading that has been posted in several decades of blogging.