Chapter 5 of 7  ยท  Catching What AI Misses

๐Ÿ” The Continuity Editor

The disciplined practice of catching, correcting, and documenting the consistency gaps AI routinely introduces.

๐Ÿ“š The continuity editor reads every page, catching what the author lost track of โ€” eye color that changed between chapters, a subplot that was never resolved. AI is a prolific but forgetful co-author. You are the continuity editor.

Overview

Every serious author depends on a continuity editor. This is the person who catches that a character had blue eyes in Chapter 2 and brown eyes in Chapter 14. Who notices a subplot introduced on page 80 was never resolved. Who flags a farmhouse that quietly became a townhouse in Act Three. Continuity errors break the reader's spell and signal that the author lost track of their own story.

AI is a prolific but forgetful co-author. It will write confidently, produce working code, and move on without looking back. The Snap AI Author is the continuity editor โ€” reading the full manuscript, catching the drift, enforcing the discipline the AI cannot enforce on itself.

Five failure patterns surface repeatedly: Incomplete rewrites (AI updates 30 of 100 endpoints and stops โ€” provide a full manifest and confirm every item). Missed abstraction (AI solves locally instead of recognizing shared patterns โ€” explicitly ask before accepting). DRY violations (duplicate logic compounds with every revision โ€” treat deduplication as a continuity check). Repeated refinement passes (one prompt rarely produces production-ready code โ€” build the re-prompt loop before every PR). Forgotten processes (AI has no memory โ€” write every refined process as a living .md file, committed to the repo, kept current, and handed to the AI at the start of each related session).

Core Concepts

  • Incomplete Rewrites: Give AI a full manifest for every bulk change and confirm every item was updated. Never assume it tracked the full scope.
  • Missed Abstraction & DRY: Before accepting generated code, ask if any logic can be shared. Treat deduplication as a continuity check, not a later optimization.
  • The Refinement Loop: Run a production-readiness re-prompt before every PR โ€” error handling, logging, edge cases, security. First draft is never final.
  • Living .md Documentation: Write every refined process to a .md file in the repo. Keep it current. Hand it to AI as context at the start of each related session.

Key Principle

AI is a prolific but forgetful co-author. The Snap AI Author is the continuity editor โ€” cross-referencing Chapter 14 against Chapter 2, catching the drift, and enforcing the discipline the AI cannot enforce on itself.