A forensic account of how fluent AI interaction can fracture without warning.
This book introduces a structured vocabulary for a rupture that was felt before it was named.
The Wrench documents epistemic drift in AI-assisted work-the moment when confident, coherent systems begin distorting judgment without triggering an error message.
While most discussions of AI risk speculate about the future, this record examines what has already occurred. It traces how trust erodes in real time, how fluency masks structural failure, and how relational interfaces obscure mechanical limits.
This is not a productivity guide.
It is not prompt engineering advice.
It is a forensic manual.
Who This Book Is For
Engineers & Architects - Designing or integrating LLM-based systems into consequential workflows.
Decision-Makers - Delegating high-stakes cognitive labor to autonomous systems.
Researchers - Working in human-computer interaction, cognitive science, and AI reliability.
Regulators & Policy Professionals - Evaluating system accountability, transparency, and epistemic integrity.
Core Themes
Recognizing when a system is confidently wrong
Detecting subtle fractures in judgment and trust
Identifying structural failure masked by fluency
Establishing frameworks for monitoring epistemic integrity
This is an empirical record of interaction breakdowns.
All claims are falsifiable.
Authority rests with the reader.