LLM Wiki 101

Definition

An LLM Wiki is a source-backed Markdown wiki that an AI helps maintain. The important word is maintain: each source should improve a durable set of pages rather than produce another disposable chat answer.

• Raw source material stays separate from finished wiki pages
• Concept pages collect reusable ideas across many sources
• Open questions hold claims that are not supported well enough yet

The folder model

Instead of asking an AI for one-off answers, you keep raw material in one place and compile durable wiki pages that can be searched, linked, and revised.

• raw/ keeps articles, transcripts, papers, and original notes
• wiki/ holds source notes, concept pages, entity pages, syntheses, and indexes
• schema/ defines frontmatter, citation rules, confidence labels, and page types
• prompts/ stores repeatable ingest, query, lint, and deepen-topic prompts

The workflow

A useful personal wiki repeats three operations: ingest new sources, query the current wiki, and lint for weak links, unsupported claims, and outdated pages.

• Ingest turns a source into source-backed Markdown pages
• Query uses the current wiki as context instead of starting from scratch
• Lint keeps quality from drifting as pages accumulate

The beginner mistake

Most people start by asking an AI to summarize a source. That creates useful text, but not a system. Start with page types, source rules, and links so every output has somewhere to live.