# Faultprint Synthesis Pass 001

## Plain-English Read

The core library is now large enough to support pattern synthesis, but it should still be treated as a structured research asset rather than a finished benchmark. The strongest commercial path is to turn repeated structural conditions into named assessment modules, then validate each module against source freshness, sector balance, and counterexamples.

## Dataset Grain

- Rows analyzed: 680
- Institutions represented: 136
- Sectors represented: 107
- AI-related rows: 390 (57.4%)

## Core Distribution

### Resemblance Bands
- moderate_resemblance: 334
- elevated_resemblance: 188
- low_resemblance: 156
- high_resemblance: 2

### Client Status
- client_ready_with_caveat: 254
- client_ready: 241
- hold_internal: 154
- revise_internal: 31

## Source Hardening Status

- P0 broken-source repairs: closed
- P1 manual/source replacement repairs: closed
- P2 retrieval metadata capture: closed
- Rows with source freshness status: 233
- Rows with retrieval date metadata: 197
- Rows with snapshot status metadata: 197
- Rows with source repair status: 232
- Rows with exact source publication date captured: 15

## Leading Pathology Signals

|pathology|row_count|client_usable_rows|elevated_or_high_rows|ai_rows|sector_count|product_signal_score|
|---|---|---|---|---|---|---|
|AI Authority Acceleration|210|203|11|210|52|561.07|
|Contestability Lag|249|126|82|87|99|553.97|
|Record-Authority Capture|70|55|38|42|25|226.87|
|Access Interruption Before Review|44|35|36|12|17|159.49|
|Documentation Burden Transfer|26|20|16|14|19|100.75|
|Classification Gatekeeping|27|26|0|14|14|71.09|
|Evidence-Container Illusion|12|12|7|4|10|51.99|
|Multi-Actor Handoff Diffusion|10|10|0|6|6|31.53|
|Deadline Compression|8|8|0|0|6|23.49|
|Provisional Relief Reversal|24|0|0|1|8|14.88|

## Strongest Co-Occurrence Pairs

|pathology_a|pathology_b|cooccurrence_count|share_of_smaller_pathology|
|---|---|---|---|
|Documentation Burden Transfer|Record-Authority Capture|529|0.9232|
|Contestability Lag|Record-Authority Capture|401|0.957|
|Contestability Lag|Documentation Burden Transfer|328|0.7828|
|AI Authority Acceleration|Record-Authority Capture|327|0.8838|
|AI Authority Acceleration|Documentation Burden Transfer|320|0.8649|
|Access Interruption Before Review|Record-Authority Capture|169|0.9657|
|AI Authority Acceleration|Contestability Lag|169|0.4568|
|Access Interruption Before Review|Documentation Burden Transfer|135|0.7714|
|Access Interruption Before Review|Contestability Lag|107|0.6114|
|Record-Authority Capture|Remedy Path Split|94|0.9038|
|Documentation Burden Transfer|Evidence-Container Illusion|91|0.8505|
|Evidence-Container Illusion|Record-Authority Capture|89|0.8318|

## Sector Signal Matrix

|sector|row_count|institution_count|client_usable_rows|elevated_or_high_rows|ai_rows|
|---|---|---|---|---|---|
|Insurance telematics|46|9|29|16|31|
|Credit / financial AI|15|3|12|6|15|
|Employment screening / consumer reporting|10|2|8|6|0|
|Tenant screening|10|2|7|5|3|
|Education finance / federal student aid|10|2|8|5|1|
|AI lending / credit underwriting|10|2|8|4|10|
|Bank account screening|5|1|4|4|0|
|CPS intake and investigation|5|1|4|4|1|
|Marketplace seller enforcement|5|1|4|4|0|
|Credit reporting disputes|5|1|4|4|0|
|Driver and vehicle reinstatement|5|1|4|4|3|
|Healthcare documentation AI|5|1|4|4|5|

## Failure Track Candidates

|track|row_count|client_usable_rows|elevated_or_high_rows|ai_rows|sector_count|
|---|---|---|---|---|---|
|Administrative Debt Triad|313|189|120|125|102|
|No named triad|132|100|36|45|66|
|AI Authority Triad|207|185|29|207|66|
|Vendor Diffusion Triad|13|12|4|8|9|
|Access Shock Triad|4|4|1|0|4|
|Review Theater Triad|11|5|0|5|11|

## Highest-Priority Institutions For Deeper Synthesis

|institution|sector|row_count|client_usable_rows|elevated_or_high_rows|pathology_count|ai_rows|synthesis_priority_score|
|---|---|---|---|---|---|---|---|
|Citizens Property Claims|Property insurance / insurer of last resort|5|4|4|3|0|43.5|
|Tesla Full Self-Driving Supervised|Driver assistance / consumer vehicles|5|4|2|5|2|42.5|
|Chase Account Restriction and Dispute|Banking / deposit accounts|5|4|4|1|0|41.5|
|Xfinity Service Suspension and Dispute|Broadband and residential communications|5|4|4|1|0|41.5|
|The Work Number Employment Data Dispute|Employment and income verification / consumer reporting|5|4|4|1|0|41.5|
|E-ZPass Toll Violation Dispute|Tolling / mobility debt|5|4|4|1|0|41.5|
|Con Edison Payment Agreement and Termination|Electric gas and steam utility|5|4|4|3|2|40.5|
|MTA Transit Adjudication Bureau|Transit enforcement / fare violation adjudication|5|4|4|1|0|40.5|
|Verizon Service Suspension and Bill Dispute|Telecom / wireless and Fios|5|4|3|1|0|40.5|
|HireRight Background Check Dispute|Employment screening / consumer reporting|5|4|3|1|0|40.5|
|GEICO DriveEasy|Insurance telematics|6|5|2|3|2|40.5|
|Microsoft / Nuance|Healthcare documentation AI|5|4|4|2|5|40.0|
|Emergency Rental Assistance Documentation|Rental assistance / eviction prevention|5|4|4|2|0|40.0|
|Lemonade|Insurance|5|5|3|4|3|40.0|
|School Discipline Due Process|K-12 education / student discipline|5|4|3|2|0|40.0|

## What This Means

1. The data is no longer just a pile of extractions. It is starting to show repeatable condition families: contestability breakdown, AI/automation authority, record-authority capture, access interruption, documentation burden, and provisional-relief reversal.
2. The commercially interesting move is not to claim prediction. It is to say: documents can be compared against structural condition families observed across prior failures and current operating environments.
3. The strongest first products are a diagnostic assessment, a sector-specific failure-track brief, and a pathology atlas. A benchmark product comes later after source freshness, sampling balance, and external validation are stronger.

## Data Quality Risks

- Exact source publication dates remain sparse. P0-P2 source hardening is closed, but broad benchmark claims still need a full archival/snapshot regime for the remaining optional P3 sources.
- Row counts do not equal institution counts. Sectors with more source documents can appear stronger unless normalized.
- Co-occurrence is row-based, not time-based. It shows conditions traveling together, not sequence or causality.
- Client-usable rows are suitable for bounded findings, not broad market claims.

## Next Build Step

Build the pathology atlas as the first real product layer: one page per condition family with definition, observable document signals, evidence examples, falsifiers, sectors, and limits. That becomes the foundation for assessment output, sales language, and later benchmarking.