RETRIEVAL FORENSICS Investigating Compression Damage in the AI Retrieval Layer

Retrieval Forensics is a methodological practice defined by Rex Fraction for investigating distortion in AI-generated entity representations. The practice focuses on the path by which meaning is altered rather than simply counting whether an entity is mentioned.

A forensic audit uses the Encyclotron across five levels: entity recognition, competitive position, intellectual-property attribution, customer-decision framing, and founder identity. Returned answers are examined for genericness, invention, omission, and fragmentation. The resulting compression map is meant to show where an entity collision, attribution scar, category erasure, philosophy loss, or decision-layer recoding occurred.

The document’s demonstration concerns Basecamp and reports that the product, Shape Up methodology, and founder entity remain separately retrievable but fail to cohere in commercial queries. It also reports unfavorable comparison framing and omission of the company’s intentional-simplicity philosophy. Retrieval Forensics is positioned as the investigative stage before constructive Retrieval Architecture and is complemented by Compression Diagnostics, Entity Integrity, and metadata-packet deployment.