Arguments and sources

Which arguments appear

Thirteen argument types coded present or absent per post. A post can use several, so shares do not sum to 100%.

What the discourse cites

Tiers are assigned by reference kind (arXiv and DOI are academic, Amazon ASINs are books, Wikipedia is encyclopedia, Substack posts are Substack) and then by domain list; images, link shorteners and platform pages are dropped. The full mapping is in config/domain_tiers.yaml. Links that point at one work under different URLs (a publisher page, its PDF and its DOI; a preprint and the journal version; a lab's blog post and its paper) are merged: distinct links became works.

Top works by tier

A work named in grey had no title we could recover and shows its URL slug, or the URL itself, instead; hover a name for where the title came from. "+n" after a site means n other URLs were merged into the work; hover to see them.

Who gets named

Counted by regular expression over post text; a post counts once per person. Surnames alone are used only where unambiguous. Names in blue link to the person's Wikipedia article; a name in grey has none. The list is in config/names.yaml.