# Condition Crosswalk Engine Spec 001

Status: first build specification

Purpose:
Compare structural conditions found in operating corpora against structural conditions found in postmortem corpora.

The two libraries are not the product by themselves. The intelligence comes from the crosswalk between them.

## Input Libraries

1. Postmortem Condition Library
   - Known breakdowns, investigations, litigation, audits, enforcement records, inquiry reports, official reviews.
   - Captures structural conditions visible in cases where failure occurred.

2. Operating Condition Library
   - Current institutional documents not framed around a known failure.
   - Captures structural conditions visible in active institutional operation.

## Unit Of Comparison

The unit is:

`condition + companion conditions + controls + proof gaps + mode`

Not:

`company looks like failed company`

## Core Questions

The crosswalk must answer:

- Which operating conditions also appear in postmortem cases?
- Which conditions are common in operating institutions and not meaningful by themselves?
- Which companion conditions make a structural condition more serious?
- Which visible controls distinguish ordinary operating exposure from postmortem breakdown?
- Which evidence gaps recur before escalation?
- Which operating corpora are outliers against the operating baseline?
- Which operating corpora resemble postmortem anatomy without enough evidence to classify failure?

## First Crosswalk Metrics

1. Condition Overlap Rate
   - How often an operating condition appears in postmortem cases.

2. Operating Baseline Frequency
   - How often a condition appears in operating corpora.

3. Escalation Companion Count
   - Number of known escalation companions present in the operating corpus.

4. Control Visibility Score
   - Whether controls are visible, asserted, evidenced, or absent.

5. Proof Gap Score
   - Whether the corpus lacks the records required to verify controls.

6. Postmortem Resemblance Band
   - Low, moderate, elevated, high.

7. Operating Outlier Band
   - Ordinary, watch, elevated, outlier.

## Output Language

Allowed:

- "This condition appears in both operating and postmortem corpora."
- "The reviewed corpus shows the condition but not the escalation companions."
- "The control is asserted but not evidenced in the reviewed corpus."
- "Compared with the operating baseline, this corpus shows elevated automation proximity."

Forbidden:

- "This company will fail."
- "This is the next Horizon."
- "The institution should..."
- "The condition proves misconduct."

## Minimum Useful Dataset

Operating Library targets:

- Phase 1: 12 operating runs to test schema.
- Phase 2: 60 operating runs for first baseline.
- Phase 3: 100 operating runs for stronger comparison against the existing postmortem corpus.

## First Domains

- insurance
- banking
- fintech
- healthcare
- AI hiring
- education
- public benefits
- utilities
- logistics
- transportation
- government digital services
- housing
- criminal justice technology
- child welfare
- emergency management