Treat every note link as infrastructure for AI graph traversal
When building knowledge systems that will interface with AI, treat every link you create as infrastructure that future graph traversal algorithms will follow, prioritizing explicit relationship encoding over implicit semantic similarity because GraphRAG systems require edges to perform multi-hop reasoning.
Why This Is a Rule
GraphRAG systems — the next generation of AI-powered knowledge retrieval — traverse your knowledge graph by following explicit edges between notes. They can hop from note A to note B to note C, building multi-step reasoning chains. But they can only follow edges that exist. If two deeply related notes aren't linked, the graph traversal hits a dead end and the AI loses access to the connection.
Semantic similarity (embedding-based retrieval) can find notes about similar topics, but it can't reason about relationships. It can tell you "these notes are about similar things" but not "this note contradicts that note" or "this note is an example of that principle." Multi-hop reasoning requires typed, explicit edges: "derives_from," "contradicts," "exemplifies," "enables."
This rule shifts your link-creation mindset from "helpful for me right now" to "infrastructure for future traversal." Every link you create — or fail to create — shapes what reasoning paths AI systems can follow through your knowledge.
When This Fires
- Creating links between notes in any PKM tool (Obsidian, Notion, Roam, LogSeq)
- Deciding whether a connection between two ideas is "worth linking"
- Building or maintaining a knowledge graph that AI systems will query
- Designing note templates or schemas for a new knowledge base
Common Failure Mode
Relying on implicit semantic similarity and assuming AI will "figure out" the connections. You have notes about cognitive load, about attention, and about deep work — all semantically related. But without explicit edges encoding how they relate (cognitive load limits attention span, which determines deep work session length), AI retrieval returns all three as equally relevant context rather than constructing the causal chain you need.
The Protocol
When creating or reviewing a note, ask: "If a graph traversal algorithm landed on this note, which other notes should it be able to reach, and what is the relationship?" Encode each answer as an explicit, typed link. Prefer relationship-labeled links ("contradicts: [[X]]", "derives_from: [[Y]]") over bare links ("[[X]]") — the relationship type is the edge label that enables multi-hop reasoning.