source-summary
Cairn - Redefining MOC for AI-Human Knowledge Structures
Cairn: Redefining MOC for AI-Human Knowledge Structures
Cairn redefines the PKM concept of Map of Content (MOC) as a three-layer system: atoms (individual authored insights), L1 MOCs (topic-level structures), and L2 MOCs (theme-level community structures). Their LLM "Moss" drafts these structures while humans retain editorial control, creating an AI-maintained, human-audited "single source of structure" that preserves the shape of disagreement rather than flattening it into consensus. The team is currently validating whether this structured preservation of multi-perspective disagreement can materially improve decision quality compared to conventional search and summarization tools.
Key Takeaways
- MOC is restructured into three tiers: atoms (authored insights), L1 topic-level structures, and L2 theme-level community structures.
- An LLM named Moss drafts MOC structures, but humans retain final authority over how knowledge is organized and connected.
- The system intentionally preserves the 'shape of disagreement' across communities instead of forcing consensus.
- For individuals, it parallels Karpathy's LLM Wiki pipeline (Schema → Wiki → Raw); for teams, it improves on Confluence via AI-maintained structure with human review.
- Core thesis focuses on validating when preserving disagreement meaningfully improves decision quality, beyond aesthetic or romantic appeals to multi-perspectivism.
- Early evals of MOC-gated retrieval show structured-answer advantages over ChatGPT and Perplexity.
Related Concepts
Related Entities
Related: overview.