Attention screens find loud; the shape that paid was quiet: add a bottom-up board judged against each entity's own past and label dead communities

A multi-source attention system had a dozen boards that all ranked entities by how much louder than usual they were. Reviewing the three cases that actually worked for the operator showed the opposite shape: small entities with no broad attention, whose own community was slowly tightening (more posts from the same people, more scarcity language). None of them ever reached the top of a board, so the system could not put the right kind of candidate in front of the operator. A second failure: the universe builder added communities that merely existed (newest post six years old, one private since spring) and every board read their silence as a quiet week.

(1) Add one board that screens for the quiet shape: every entity under a size cap with a LIVE owned community, judged only against its own past on a few independent reads (community volume, recent weeks vs the prior twelve; outside mentions per 1k of the corpus; the community's own scarcity/reseller share with its lift; next dated catalyst; days on the other boards). Sort live-first, then by community tightening; print no ratio when the base is thin. Say explicitly that it is observation, not prediction. (2) Define community liveness from data, not existence: require N posts in the last 60 days at build time, and on the board show "no post for K days, not a community now" instead of a blank. Verify suspicious silences in a real browser: the archive was right in every case (private/dead), the resolver was wrong to trust existence. (3) Persist the board daily so it can be scored later against the operator's own logged calls.

Ranking by loudness selects for what everyone already sees; the exploitable cases were the ones nobody was looking at, so the screen must start from the population (all small entities with a community) rather than from the noise. Existence of a community is not evidence about the entity; its recent activity is.