nx_benchmark_suite_registry.nx

nx_benchmark_suite_registry.nx -- the RED-TEAM META-HONESTY organ: censuses our own CENSUSES. For each workstream it names the EXTERNAL benchmark SUITE (like MSU for video, Unsloth for LLM-training) and marks whether we are WIRED to it (ran the real tool h2h, both numbers) or UNWIRED (self-graded only = our score is SUSPECT). Operator 2026-07-10: "get suites of tools up for benchmarking so we stop giving ourselves super positive ratings when the nishi red team and critic ... should be calling out our shit." This organ deflates inflated self-ratings BY CONSTRUCTION: an UNWIRED workstream's high self-score is printed as UNVERIFIED, and the "self-congratulation risk" = the count of high self-scores with no external measurement. NOT a self-gate that says we're great -- a self-gate that says WHERE WE HAVE NO RIGHT TO CLAIM ANYTHING YET. license_tier: ORIGINAL