# AI Coding Defaults to Addition Over Maintenance ## Claim AI coding defaults to adding locally satisfactory code instead of improving the shared structures a codebase must maintain. <!-- atomicity: 1 idea ✓ --> ## Scope & conditions - This applies where prompts reward immediate completion without reuse-first review, legacy-care requirements, or cause-based error handling. - The reported association doesn't prove every AI-assisted commit causes decay; team practices and model choice can change the result. - The default can be countered by checks for duplication, refactoring, connectivity, and swallowed errors. ## Evidence [[3. Resources/Readwise/Articles/Code Maintainability Plummets in the AI Coding Era]] reports duplication up 81%, legacy refactoring down 74%, and more silent error masking across 623 million changes from 2023–2026. ## Links - [[Cheap Generation Raises the Total Maintenance Bill]] — evidence-for: generated additions expand the future maintenance surface while legacy care falls. - [[Clean Code Is the Human-AI Interface]] — contrasts: shared abstractions and explicit error paths are the structures the default generation pattern erodes. - [[The Bottleneck Moved From Code Generation to Review and Judgment]] — applies-in: reviewers must catch locally plausible duplication and masked failures before they become system debt. - [[3. Resources/MOCs/AI & the Future of Knowledge Work MOC]] — applies-in: the MOC lists this note under Maintenance & durability.