Federal and state case law, legal briefs, judicial opinions. Detects accountability theater and AI-fabricated reasoning before it reaches the bench. The validation corpus includes the Mata v. Avianca fabricated brief — scored 7/100, the largest single-document gap in the corpus.
10-K, 10-Q, 8-K, proxy statements. Built on the SEC EDGAR public corpus. SVB's December 2022 risk disclosures scored 22 — three months before the March 2023 collapse. Enron FY2000: 8/100. The signal was in the structure before the failure became public.
Bills, committee testimony, presidential addresses, Federal Register notices. Calibrated against the founding documents — Federalist No. 51 (91), Gettysburg Address (89). The Iraq WMD Senate hearing scored 21. The standard has always existed. Now it is measurable.
One engine across all three verticals
The four-gate scoring architecture (G4 Paradox Resolution · G6 Latent Intent · G7 Argumentative Structure · G8 Rhetorical Architecture) is identical across legal, business, and government documents. A T2 (Functional) is a T2 whether it comes from a federal brief, a 10-K, or congressional testimony. Channel-capacity measurement is domain-agnostic; the documents are domain-specific. See How It Works for the engine architecture, or Research for the theoretical foundation.