nx_adopt.nx

nx_adopt.nx -- THE ADOPTION CENSUS. Which capability primitives has nobody adopted? (operator priority) LAW: THE BOTTLENECK IS NOT BUILDING PRIMITIVES, IT IS ADOPTING THEM. Three proofs on 2026-07-30: cpu-feature-detect existed with a KAT, registered nowhere (seq1359) - ss_close: 133 openers, 1 closer (seq1347) - nx_hw_cpu_count: CORRECT (sched_getaffinity, cgroup-safe) with ZERO external callers in 19,823 files while nx_torrent_get hardcodes MAXP=32 (seq1410). A build-only 'done' is the cause. ==> A ZERO-CALLER CAPABILITY PRIMITIVE IS A RED DEFECT, NOT A COMPLETED FEATURE. SCALE: the naive census greps the corpus once per primitive = O(primitives x corpus). This walks the corpus TWICE total -- pass A hashes every definition, pass B counts every identifier against it with O(1) lookups -- so cost is O(corpus) regardless of how many primitives exist. SELF-REFERENCES DO NOT COUNT AS ADOPTION. nx_hw_cpu_count showed FIVE grep matches and looked alive; four were its own file. Definition-file hits are tallied SEPARATELY from external ones -- that is the entire difference between 'used' and 'looks used'. argv: <defs-dir> [callers-dir] exit 0 = ran, 2 = usage/unreadable. license_tier: ORIGINAL Read-only. No hw writes (Rule 26).