FINQUILL://methodology

How Node produces briefs.

Every Audit Whisperer brief, governance brief, parameter interpretation, and composability cascade alert passes through a multi-pass LLM editorial pipeline before it's published. We're betting that verifiable reasoning + published accuracywins over opaque recommendation. Here's the chain every brief runs through.

[01]

Synthesis

A primary LLM (Claude Opus) reads the source — audit report PDF, governance proposal text, parameter-change event — and produces a structured brief with explicit evidence links to each claim.

[02]

Evidence completeness

A validator LLM checks every claim in the brief. Each must trace to a citation in the evidence list. Unverified claims block publication.

[03]

Adversarial critique

An independent LLM persona (different system prompt, sometimes different model) tries to find errors, contradictions, and missing context. Critical findings requeue the brief for re-synthesis.

[04]

Fact-check

A validator LLM with retrieval over the source data re-fetches each cited URL and verifies the verbatim excerpt matches. Mismatches block publication.

[05]

Severity guard

A heuristic + LLM pass. Findings labelled 'critical' or 'definitive' require ≥2 independent citations. The bar is higher for higher-impact claims.

[06]

Publication gate

All checks pass → editorial_status flips to 'published' and the brief becomes visible. Any failure routes to a review queue for manual investigation.

// accuracy track record

Published accuracy. Quarterly.

Predictions made in Node briefs (governance outcomes, parameter consequences, audit-finding severity) are tracked inaccuracy_tracking. Resolved predictions feed a public quarterly scorecard. The first report ships Q4 2026. Methodology is disclosed. Wrong claims show up in the data.

// Predictions tracked
0
first quarter Q4 2026
// Published briefs
visible at /audit
// Confident-wrongness
0
we report misses publicly
// who we're NOT

What Node is — and isn't.

  • Node briefs are analysis, not investment advice. Confirm against the original source before acting.
  • We synthesize publicly available material. Audit firm reports, governance forums, on-chain events. Never proprietary data.
  • Position-aware filtering and per-user briefs are credit-metered. Reading the public Audit Whisperer is always free.
  • We publish misses. The accuracy track record is the trust signal.