D128 - 30,610 names mined from seven emulators. The hash reproduces 30,137 of them, and every exception is the source's naming, not the hash

Status: derived - by comparison against every identifier the toolkit publishes.

scripts/mine-nids.py reads seven emulators in six formats and produces data/mined-names.txt: every symbol name any loader in the toolkit knows, with the identifiers it claims.

source	rows
fpPS4 ps4libdoc.pas	78,372
shadPS4 LIB_FUNCTION	5,174
PS5PCEM Zig tables	1,496
SharpEMU	1,190
craziiEmu	945
GPCS4	899
ChonkyStation4	507
ps4libdoc known_names	22,382

30,610 unique names, 22,944 corroborated by two or more independent sources. obSCEne knows 564 and 530 of those appear here, so the surface would grow 54-fold.

The hash, checked against all of it

30,137 match. 473 do not, and none is a hash failure:

• 223 are placeholders. Func_00F4D778F1C88CB3 and _import_060337B772EF70D9 are fpPS4's way of writing "the function whose identifier is this" - the hex is in the name. Not names at all, and a free source of unnamed identifiers.
• The rest are implementation names, not symbol names. SysmoduleGetModuleInfoForUnwind hashes to 0x0563ea3d... against a claimed 0xe1f539ca...; restore the prefix and sceSysmoduleGetModuleInfoForUnwind hashes to 0xe1f539caf3a4546e exactly. shadPS4 names its C functions for the reader, and one C function can serve several symbols - sceHmdDistortionGet2dVrCommand is registered under two identifiers in two libraries, only one of which is that name's hash.

So the corroboration is now seven-way and spans both generations.

Two things that must happen before any of this is ingested

A name that does not hash to a claimed identifier must not enter the surface. It would make obSCEne import a symbol that cannot exist - harmless, since an unresolved import is reported absent, but 250 rows of noise in a census whose value is that its absences mean something. The miner needs to check each row against the hash and mark it.

25,320 of the 30,610 have no library attribution, because only three of the seven sources publish it. That matters for obSCEne's own answer and not at all for supplying identifiers to a name search - the two goals want different things from the same corpus.

The number that reframes the exercise

ps4libdoc carries 1,130,742 identifiers observed in real modules with no known name, against the 30,610 names collected here. 2.64%.

Mining every emulator in existence still leaves 97% of the observed surface nameless. The named set is the rounding error; the unnamed identifiers are the interesting half, and obSCEne can import those directly because the import is the identifier and the name only ever existed to compute it.