A reference desk for knowledge that models can reuse carefully.

LLM Wikibase exists for the operational side of AI knowledge: the part between raw documents and final answers. Many teams have documents, glossaries, and prompts, but fewer have a shared layer that names entities consistently, records boundaries, explains evidence, and prepares compact notes for retrieval.

This site treats that layer as a public craft. It studies how records should be shaped, how retrieval lanes should be separated, and how review signals should travel with a claim. The tone is practical and database-minded: useful records should be inspectable, boring in the right places, and precise enough to reduce unnecessary interpretation.