Patlas by will

If you keep dozens of papers open "to read later", Patlas turns that pile into a library you can actually search.

It watches the tabs you already have open, recognises which of them are academic — arXiv, OpenReview, PDFs, conference proceedings, publisher pages, and the project pages that have no tell-tale URL at all — and files each one with its title, authors, year, venue and abstract.

TWO WAYS TO FIND THINGS AGAIN
Every paper gets any number of categories (a paper can be both goal-conditioned RL and a diffusion method, and Patlas records both) plus overlapping keyword labels. Filters compose. Drag a card onto a category to file it there too.

IT KNOWS WHAT IT DOES NOT KNOW
When two records might be the same paper but the identifiers disagree, Patlas keeps both and flags the pair for you to decide, instead of silently guessing. Papers sharing a DOI or arXiv id are merged automatically.

BUILT FOR A BIG TAB PILE
The periodic sweep never wakes a tab your browser has unloaded, so it does not undo the memory the browser just reclaimed. "Close saved tabs" closes only the ones whose papers are confirmed stored — pinned tabs, tabs playing audio, and anything still needing review stay open.

YOURS, LOCALLY
The library lives in your browser. No account, no server, no analytics. Export it to JSON whenever you want. The only network requests are metadata lookups to arXiv and Crossref, and you can switch those off in one click.

Favourites, a trash can with restore, and per-paper screenshots for pages that publish no abstract.