If you had a bunch of collections of
2026-08-18 01:35:02.534272+02 by Dan Lyke 1 comments
If you had a bunch of collections of information, perhaps envisioned by folders full of PDFs and web links, how might you lay out clusters of that?
I can use semantic embeddings, but the two largest eigenvalues doesn't seem like enough info.
Also, what's the right granularity for those embeddings? More than folder name, I'd guess.
Maybe something involving distances in that space between each of the collections? What's the right algorithm for reducing that to 2D?