nx_ivvcensus.nx

nx_ivvcensus.nx -- derive the IV+V registry from the WHOLE estate instead of a hand-picked sample. WHY. nx_ivvbeat shipped verifying FOUR gates. buildroot/runtime holds ~1,850 *_gate.nx organs and knowledge/status holds 363 verdict logs, 233 of them carrying a judgeable anchor. Four rows is 0.2% coverage that nobody ever justified -- the registry was a sample masquerading as a roster. (STAR)A HAND-WRITTEN ROSTER IS A SAMPLE WHOSE SIZE NOBODY CHOSE; IT IS NOT SMALL BECAUSE SOMEONE DECIDED IT SHOULD BE, IT IS SMALL BECAUSE SOMEONE STOPPED TYPING. METHOD -- MATCH A KNOWN GOOD, THEN GENERALISE. The marker for each gate is not invented here; it is read out of that gate's OWN log, which is the only place its success string is recorded as fact. Where the log shows a "passed N/M" shape the HIGHEST M ever observed becomes the tooth floor, so a gate is held to the best it has ever demonstrated rather than to a number someone typed once. (STAR)THE STRONGEST FLOOR YOU CAN SET WITHOUT GUESSING IS THE BEST THE THING HAS ALREADY PROVEN. SCOPE DISCIPLINE -- THIS PROPOSES, IT DOES NOT SEED. Every candidate row is written to a proposal file and nothing touches the live ivvreg- plane. A census that silently rewrote the registry would make the beat verify a roster nobody had read. NEVER-BRICK (Rule 26, ABSOLUTE and NOT waivable by any blast-radius instruction): this organ EXECUTES NOTHING. It reads logs and tests for file existence. The firmware-axis gates (uefi/boot/efi/cmos/fw) are additionally flagged so that whoever seeds the registry must decide about them deliberately rather than inherit them in a bulk paste. license_tier: ORIGINAL No hw writes. Executes nothing. Writes ONE proposal file.