3. Verifying the NID mapping - done, and far better than hoped

obscene-tool crack exists (D049), data/nid-corpus.txt holds 389 pairs harvested from an emulator's own resolution log, and the hash chain has been checked against a large external database.

78,372 of 78,372 pairs agree (D063). Those identifiers were extracted from real hardware firmware rather than computed, so reproducing them is agreement with hardware values across a large sample - it rules out every byte-order, alphabet and digest-truncation mistake at once.

With one correction. The two databases used are not independent of each other: fpPS4's table is a superset of ps4libdoc, sharing 42,009 of its 42,010 names. Presenting them as two witnesses was wrong (D064). The honest count of independent confirmations is two: the published test vector, and this database.

What is still absent by choice is a candidate generator. Describing the naming convention is the actual problem and it wants iterating on, so it belongs in a script that emits a word list rather than compiled into the tool.