Everything below is genuine output from the debate engine behind BoardCouncil, run on 14 July 2026 against a synthetic board pack. The company and its figures are fictional. The argument — and the verdict — are not.
We ran a second, subtly different claim through the same engine: "the Q3 board pack states 7 months of cash runway." That claim was upheld — because it's literally true: the pack does state it. This is exactly what the evidence grades are for. That the pack says 7 months is Observed. That the figure is supported is Refuted. A single-analyst summary blurs the two; a director who repeats the first as if it were the second carries the difference into the boardroom.
The adversarial core is real and it discriminates: attacked, defended, and adjudicated with the correct financial reasoning — recomputing burn from the balance sheet instead of accepting the deck's framing.
A production pre-meeting brief adjudicates many claims from a full pack, registers cross-document conflicts, carries calibrated confidence, and seals the whole run into a hash-chained audit record. This page is one claim, shown end-to-end so you can judge the mechanism.
The complete path — intake, enrichment, conflict register, judgment, your gate, the brief — is on How it works. Or bring one real pack to the pilot and see it on your own materials.
Bring one real board pack under NDA; we'll return a graded, adversarially-tested pre-meeting brief — and you'll see the audit trail behind every finding.