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.