buildability
Judge-run comparison test that measures disagreement between independent evaluators on the same Buildability Receipt.
Adversarial Divergence is a robustness check. Multiple independent LLM judges and a human reviewer score the same Buildability Receipt and the platform records how often they disagree. Low divergence means the assessment is stable; high divergence flags a check that is ambiguous. Divergence is a Wave 7 completion gate and the outcome is itself receipt-backed.
Source: curated glossary catalog. Freshness: git_versioned_curated_catalog.
Term API · apps/web/data/glossary/terms.ts