llm-wiki

A persistent, file-based knowledge base for AI-assisted software delivery, implementing Andrej Karpathy's "LLM wiki" pattern: knowledge is compiled once into an interlinked markdown wiki and kept current by an agent — rather than re-derived from scratch on every query (RAG).

Everything lives in .asdlc/knowledge/. It is plain files — no database, no server, no infrastructure. Humans read it in any editor or on GitHub; agents read and maintain it through one entry point.

Why

"The tedious part of maintaining a knowledge base is not the reading or the thinking — it's the bookkeeping." — Karpathy

That bookkeeping is exactly what an agent does well. This repo gives the agent a schema, a toolchain, and a contract so the wiki stays consistent, sourced, and trustworthy over time.