Rationalsphere Search

Paste a LessWrong URL to find semantically similar content

URL not found in corpus. Make sure it's a full LessWrong or Substack URL for a post in our database.

About the corpus

Last updated 2026-05-11 · 375,518 docs · 354,565 searchable

LessWrong — 36,598 posts, 6,843 top level shortforms, 291,051 comments. Scraped from posts with karma ≥ 5.

Substack — 41,026 posts across 173 publications, scraped from a curated list of potentially relevant blogs.

Each doc is split into ~360-token chunks (heading-aware, sliding window, or whole-doc for short pieces), embedded with Voyage-3.5, then averaged into a single doc vector. Similarity is cosine distance; top 100 matches per query. Very short docs (mostly stub comments) are kept in the corpus but not semantically indexed — those are the gap between "docs" and "searchable" above.