Module 9: Failure Modes (Why Chunks Die)
Module 9 of 9
What This Teaches
Why content disappears from AI answers.
Common failures prevent chunks from being retrieved, cited, or reused. Related: Failure Modes Documentation
Common Failures
- Pronouns: "It", "they", "this" require context that may not be preserved
- Implied context: Facts that depend on previous sentences or sections
- Mixed services: Chunks that describe multiple services or capabilities
- Marketing adjectives: Superlatives and promotional language reduce citation eligibility
- Narrative transitions: Sentences that connect ideas but contain no facts
Key Truth
These failures cause chunks to be:
- ignored during retrieval
- skipped during citation
- mutated during extraction
- replaced by competitor content
Prechunking Rule
If a chunk cannot stand alone, delete it.
Rewrite it as an atomic fact, or remove it entirely. Broken chunks are worse than no chunks.
Optional Operator Task
Task: Audit one page of your content. Identify every chunk that contains pronouns, implied context, marketing adjectives, or narrative transitions. Delete or rewrite each broken chunk.
Constraint: Every remaining chunk must be able to stand alone. If a chunk requires previous context, it fails the isolation test.
What success looks like: You produce a page where every chunk is atomic, explicit, and self-contained. No chunk dies during retrieval because no chunk depends on context that may not be preserved.
This task is optional. No submission required. No validation. Use it to convert theory into applied thinking.
Course Complete: Return to Course Overview