Progressive disclosure is the technique of navigating a knowledge bundle one directory level at a time, using index files as the map, rather than loading the whole bundle into context. It is how both agents and humans stay oriented as a wiki grows to hundreds of pages.

How it works

Each directory can carry an index.md that lists its contents — a link and one-line description per concept, grouped under headings. An agent answering a query reads the root index first, follows the relevant section index, and only then reads the specific concept pages it needs. The cross-links between concepts let it fan out from there. This works surprisingly well at moderate scale (~100 sources, hundreds of pages) and avoids the need for embedding-based retrieval infrastructure.

Why it scales

Index files are cheap to read and give the agent a content-oriented catalog of what exists. Because the bundle is graph-shaped, not just tree-shaped, the agent can also traverse relationships directly once it has a foothold. When a bundle outgrows what index files can handle, add a search tool such as qmd — but not before.

Progressive disclosure is a primary motivation for keeping index files current on every ingest; a stale index defeats the mechanism.

Citations

1. OKF README & Reference Agent
2. Karpathy — LLM Wiki gist