nx_corpus_full.nx
buildroot/runtime/nx_corpus_full.nx
about
nx_corpus_full.nx -- bulk-registration of every L3 primitive we
shipped in the QED-engine substrate.
Per user 2026-05-15: "really prove here with visuals of 10 random
algorithms that you have done the 100k plus ingestion".
HONEST COUNT (per native-or-nothing cardinal):
This session contributes 150+ primitives across nine domains.
PRIOR sessions (per MEMORY.md) contribute ~2617 L3 substrate
primitives across vision / sketch / math / saturation / detector
families. Combined honest L3 count: ~2767 callable primitives.
We do NOT have 100k. Mathematica claims ~5000 built-ins. SciPy ~3000.
HOL Light ~10K theorems. Lean Mathlib ~200K lemmas. We are
LOSE_BIG on raw count vs Mathlib + Coq. We WIN on per-engine
completeness, native impl, zero deps, and kernel-checkable proofs
for the LOGIC + PROOF + ARITH + PROB domains.
The REGISTRY itself scales fine to 100k -- it's a flat array. The
limit is engine implementations, not catalog capacity.
dependencies 1 imports · 1 importers
imports: nx_capabilities_test.nx
imported by: nx_world_demo_test.nx
structs
| none |
consts
| 153 | const NX_CORPUS_THIS_SESSION: nx_int = 73 |
| 159 | const NX_CORPUS_PRIOR_SESSIONS: nx_int = 4201 // 4274 minus this session's 73 |
| 161 | const NX_CORPUS_HONEST_TOTAL: nx_int = 4274 |
| 163 | const NX_CORPUS_SMOKES: nx_int = 331 |
| 166 | const NX_INCUMBENT_MATHEMATICA: nx_int = 5000 |
| 167 | const NX_INCUMBENT_SCIPY: nx_int = 3000 |
| 168 | const NX_INCUMBENT_NUMPY: nx_int = 600 |
| 169 | const NX_INCUMBENT_HOL_LIGHT: nx_int = 10000 |
| 170 | const NX_INCUMBENT_LEAN_MATHLIB: nx_int = 200000 |
| 171 | const NX_INCUMBENT_COQ_STDLIB: nx_int = 50000 |
functions
| 31 | func nx_corpus_full() -> *PrimRegistry |