First Reality Gap Detection.
The first proof that's a product people buy tomorrow morning.

01The killer query — show_reality_gap_report(tenant_id)

12 gaps detected

3 Shadow Authority
4 Undocumented Delegation
2 Policy Bypass
3 Responsibility Drift

money quote:  Policy ≠ Reality.

Then drill down per gap — every RealityGap carries its expected and observed state references, affected authorities / decisions / policies / assets, the responsibility chain (when present), severity, and remediation status.

02The four gap classes · v1 only (Eric LOCKED — DO NOT extend)

Eric: "Don't build twenty. Those four alone are worth a product."

03The killer demo — Expected vs Observed Organization

Eric: "Honestly this may become your strongest demo. The visual sells itself."

An internal auditor reads this in seconds: the right column has things the left column doesn't authorize. That's the entire product.

04The money quote

05The pattern — five money quotes, one corpus

06The two tenants — demo and clean baseline

07The schema — RealityGap as a first-class object

Eric: "Create a first-class object: RealityGap. You will use this object everywhere later."

RealityGap {
    gap_id,
    gap_type,                       // ShadowAuthority | UndocumentedDelegation
                                    // | PolicyBypass | ResponsibilityDrift
    severity,                       // Informational | Warning | Critical | Systemic
    detected_at_ms,
    expected_state_ref,
    observed_state_ref,
    affected_authorities: [...],
    affected_decisions: [...],
    affected_policies: [...],
    affected_assets: [...],
    responsibility_chain,           // when the gap touches a chain
    remediation_status,             // Open | Investigating | Remediated | AcceptedRisk
    explanation,                    // human-readable; surfaced verbatim to auditors
}

RealityGapReport {
    tenant_id,
    detected_at_ms,
    total_gaps,
    shadow_authority_count,
    undocumented_delegation_count,
    policy_bypass_count,
    responsibility_drift_count,
    gaps: [...],
    money_quote: "Policy ≠ Reality.",
    honest_limit_caption,
}

No new event kinds. No DB schema migration. The detector is a thin composition over existing snapshot + canonical event log inputs — the engine surface stays stable.

08Where this proof sits — rung-deepener (NOT a new rung)

Eric: "Do not create a new ladder rung. Reality Gap Detection strengthens Responsibility survives authority + Consequences survive responsibility. It's a rung-deepener. Not a rung-creator."

09What this proof IS and IS NOT

10Honest limits

1. v1 detection heuristics are LOCKED simple. Shadow Authority and Undocumented Delegation use "phantom granter / phantom delegator" matching — a principal who never received any scope. This catches the clear cases; semantic scope-derivation (e.g. delegate_review → recommend_credit) is intentionally NOT flagged. That nuance is a v2 concern.
2. Policy Bypass detects unregistered policy_ids only. A decision citing a policy_id that was registered but DEPRECATED at decision time is not yet flagged. v2 territory.
3. Responsibility Drift requires a responsibility chain. Decisions without an attached chain are not checked.
4. The detector reports facts, not judgment. A gap is a fact; whether it represents fraud, error, emergency, or undocumented business reality is determined by the Risk Committee.
5. Policy ≠ Reality. The money quote.

11Evidence appendix

12Readiness determination

What this unlocks: an Internal Audit team, CISO, or Risk Committee can now ask "What changed without authorization?" and receive a deterministic, signed, structurally-grounded answer. Engineering order Eric LOCKED: detection engine ✓ shipped here; report ✓ shipped here; dashboard is post-v1. "If the report is good enough, people will pay before the dashboard exists."

What this does not unlock: fault determinations, intent attributions, or correctness verdicts. Those remain Risk Committee / legal territory. Policy ≠ Reality.