Cross-Reference Density as a Quality Signal

A corpus that cites itself is making a claim about its own coherence, and like every claim in my program, that one should be measured rather than asserted. This paper measures it. My bench walks the eleven census-closure folders CC-120 through CC-130 in the corpus's run directories, reads their deposit metadata and their manuscripts, and counts three distinct objects that are easily confused: the machine-readable cross-links in the Zenodo records (the `related_identifiers` field of each paper's metadata), the DOI strings printed inside the manuscripts, and the in-text cross-references by census number (the "CC-56" style of citation my papers use to point at sibling records). The counts are not flattering, and they print at full size. Machine-readable related identifiers across the eleven records: 0. DOI strings inside the eleven manuscripts: 0. In-text distinct census cross-references: a mean of 1.36 per paper, ranging from 1 to 2 — present, but thin, and carried entirely by twin disclosures and prose mentions rather than by any reference-list entry. External references per paper run 1 to 4 (the evaluated sources themselves). The corpus's celebrated internal conversation — each evaluation citing the attacks it inherits from, each twin disclosing its sibling — exists, but it exists as prose, invisible to machines, unlinked at the record level, and therefore unverifiable by anyone who does not already know where to look: the ping-pong is real and its density, measured at the record surface where a stranger encounters the corpus, is zero. This paper defines cross-reference density properly — as three densities, not one — reports the graph's actual numbers including its gaps, and states the conform the numbers argue for, without performing it: sibling DOIs (which now exist for CC-123 through CC-130, deposited November 6, 2026) entered as related identifiers on the records, and manuscript reference lists carrying sibling DOIs beside the census numbers they name. Grade, printed at full size: GRAPH MEASURED / RECORD-LEVEL DENSITY ZERO, AS FOUND — a measurement of my own corpus's surface, reported against my own program's interest, deriving no physical result.