Loop Engineering

A Better Way to Think, Create, and Work with AI - Companion site

← All chapters

Chapter 12 - Failure Modes by Region

Last updated 2026-07-13. This page hosts the operator-burden literature behind Chapter 12 - the cross-disciplinary grounding for the per-region failure-mode framework - and links to the live region-by-region diagnostic prompts and the failure-diagnosis worked example.

Companion activity

Use the diagnostic prompts and worked example to classify one failure by region before choosing a fix.

What this chapter argues

When the partnership misbehaves, the question is not "is the AI broken?" - it is "which region failed?". The chapter offers a failure-pattern map (Figure 12.1) that ranks candidate regions from a symptom and names a discriminating check for each: reasoning theatre (prefrontal), loopmaxxing (working memory), missed intent (dialog interface), side effects (motor / sensory), prompt injection (motor / sensory + nervous system), context drop (working memory + long-term memory). It is a hypothesis map, not a unique localization algorithm. Each region has its own symptom, its own root cause, and its own repair. Naming the region first prevents the most expensive class of failure: throwing the wrong fix at the wrong layer.

Key sources

Chapter 12 cites in the print book: D50. Entries below are the sources this page draws on; some are related literature cited elsewhere in the book rather than in this chapter.

Primary-source links

Chapter contents

Related resources

The region-by-region diagnostic prompts page provides copy-paste-ready diagnostic probes the practitioner can run when a failure is observed but the region is not yet identified. The failure-diagnosis worked example walks one real incident through the full diagnostic process.