LLM SQUARED | COMPLETE FICTIONAL DECISION SAMPLE

Open decision-brief.pdf first. It is an illustrative proposed decision, not institution approval.

The complete supporting file is here: source-register JSON/CSV; six fictional sources; exact inputs; complete E01/E02/E03/E04 outputs; every retained test row; all exclusions and five open questions; blank decision-owner/reviewer/review-date fields. The brief cites evidence IDs; the source register maps them to exact supplied fictional versions and SHA-256 fingerprints. Fingerprints compare bytes and do not authenticate sources or people.

The supported account-service modeled USD totals exclude tax, FX, AI entitlement and exit/export charges. They are fictional amounts, not savings, verified usage or a full renewal cost. Both quote source confirmations are false, so ranking is withheld. The renewal plan is blocked. Every native pilot Pass/Fail/Invalid/Error and possible repeated attempt remains; case004 is missing, an extra row is unresolved and one row has an illustrative owned exclusion. Independent institution acceptance stays pending; zero Scorecard samples qualify.

Reproduce locally with Node22+ and no dependencies:
node methods/reproduce.mjs C:/your/new-output-folder
Use a NEW folder; this writes only regenerated outputs and makes no network request. Compare regenerated JSON to evidence files. Runtime modules are byte copies of the existing current public methods, listed in method-register.json.

No real customer, quote, review, approval, signature, current booking, achieved savings, financial action, legal opinion, certification, identity check or core integration is established. No notice, reminder or message is active. Native evaluations were not executed; the CSV is explicitly fictional.

All files are useful without a sales step. The example does not authorize installation, payment, outreach or deployment.
