nx_field_lib.nx

nx_field_lib.nx -- THE FIELD DISCOVERY CORE (/compare/fieldwatch FW1, 2026-09-05): the industry's own lists, read mechanically, become a DISCOVERED population of rivals per domain -- never a seat's six picks. WHY (operator 2026-09-05, on the modding board's six columns): "those arent a good sample of the industry ... the /compare emitter lags heavily there instead of giving a true beyond gartner or google search return alternatives and competitors and beyond real view for all these domains." A column is a choice; a field is a measurement. This lib turns three kinds of public list into candidate rows with EVIDENCE COUNTS: wiki-raw Wikipedia wikitext (action=raw): every [[Target]] and [[Target|Text]] link OUTSIDE <ref> blocks and OUTSIDE {{templates}} (citations and infobox plumbing are not rivals), namespace links excluded by their colon (Category: File: Help: Template: ...), counted per occurrence gh-topic a GitHub topic page: the repository anchors, whose markup carries the class token the page itself uses for the repo name; the two-segment href is kept as the link md-list a raw Markdown list (awesome-style): every - [Name](link) or * [Name](link) item One table holds every candidate across seeds; a name seen in several seeds carries a seeds_hit mask and a mentions count, and fl_emit writes rows sorted by seeds then mentions, so "who the field agrees on" is arithmetic. The table is a declared population: when it fills, capped=1 is ANNOUNCED, never silent. license_tier: ORIGINAL No hw writes (Rule 26).