An LLM wiki compounds where RAG rediscovers

RAG rediscovers knowledge from scratch on every question; nothing accumulates. An LLM-maintained wiki reads each source, extracts the key claims, and integrates them into an interlinked set of notes. The wiki is a persistent, compounding artifact (see: persistent memory lets corrections compound): cross-references are in place, contradictions are flagged, the synthesis reflects everything read. The human curates sources and asks questions; the LLM does the bookkeeping.

The tedious part of maintaining a wiki is the bookkeeping: updating cross-references, keeping summaries current, flagging contradictions, maintaining consistency. Humans abandon wikis because this maintenance grows faster than the value. LLMs do not get bored and can touch many files per pass, so the maintenance cost approaches zero. The human curates sources, directs analysis, and asks questions; the LLM handles everything else. The wiki becomes an asset that gets richer with every source added, rather than a burden that decays.