MirrorNet: The First Compression-Aware Engine for Narrative Fracture Detection

www.contradish.com

MirrorNet is a diagnostic engine built on the Compression-Aware Intelligence (CAI) framework that applies contradiction-compression principles to reveal hidden fractures and unresolved tensions within human narratives and identity structures. This record documents the earliest application of CAI to narrative analysis, serving as a foundational reference for future research and tools exploring contradiction-aware diagnostics.

TLDR: when a model receives two prompts that are semantically equivalent, it should produce functionally equivalent outputs. if small wording changes cause noticeably different outputs, that means the model is storing meaning in unstable or mismatched latent representations. hallucination happens when the model tries to reconcile conflicting latent states during generation.