LLM Wiki Pattern
Last updated: 2026-05-03
Summary
An LLM Wiki is a durable knowledge base where original material is stored under raw/ and synthesized pages are maintained under wiki/. The useful behavior comes from repeated distillation: an LLM reads sources, updates stable pages, and keeps claims traceable to the original inputs.
Key Points
- Raw sources and synthesized knowledge should be kept separate so the wiki can
improve without losing provenance. source
- The wiki should be navigable through indexes and related links, not just a
pile of disconnected summaries. source
- This pattern is most useful for long-lived context such as agent memory,
project knowledge, research notes, and team knowledge bases. source
Related
Sources
Open Questions
- Which source ingestion format should this repo standardize on for your daily
workflow?
- Should source distillation be manual, scripted, or driven by a specific agent
command?