# Faultprint v17 Null Pattern Sweep Summary 001

## Purpose

v17 closes Claude Closure Verification Report 003 by replacing the prior tag-based null sweep with a pattern-based null detector across the full baseline.

## Pattern Sweep Result

- Pattern-matched rows: 131
- Detector fields: condition_name + evidence_anchor + allowed_claim
- Action: all matched rows moved/kept hold_internal, low_resemblance, and null claim template.

## Client-Visible Rows Changed

- OCL-0484 / Waymo Safety Case: client_ready, moderate_resemblance -> hold_internal, low_resemblance
- OCL-0505 / Meta Ranking AI: client_ready_with_caveat, low_resemblance -> hold_internal, low_resemblance
- OCL-0510 / YouTube Recommendation System: client_ready, moderate_resemblance -> hold_internal, low_resemblance
- OCL-0515 / TikTok For You Recommendations: client_ready, elevated_resemblance -> hold_internal, low_resemblance
- OCL-0530 / Citizens Property Claims: client_ready, elevated_resemblance -> hold_internal, low_resemblance

## Client Output Status Counts After v17

- client_ready: 241
- client_ready_with_caveat: 254
- hold_internal: 154
- revise_internal: 31

## Current Baseline Files

- FAULTPRINT_OPERATING_POSTMORTEM_CROSSWALK_017.csv
- FAULTPRINT_CLIENT_OUTPUT_ELIGIBILITY_015.csv
- FAULTPRINT_V17_NULL_PATTERN_SWEEP_CANDIDATES_001.csv
- FAULTPRINT_V17_NULL_PATTERN_SWEEP_CLOSURE_001.csv

## Limit

The detector is intentionally conservative. Rows with null/absence/adjudication-not-found language are internal until a corpus-manifest treatment exists.