Essays

The arguments behind the rubric: community llms files against proprietary formats, spec v2 against v1, and the vocabulary.

• Conceptual vs proprietary llms files — Why a file no vendor owns can be trusted: the two axes, the precedence ladder that lets the most correct idea overwrite, and the rules that keep the losers visible.
• V2 vs V1 — Two versioned things share a name: the llms.txt spec (v1 → v2, 2026-08-10) and the hub pipeline (V1 site dumps → V2 acquire, refine, dual index, gate). Both diffs, a migration guide, and what breaks.
• The vocabulary file — llms-vocabulary.txt is the lexical layer of a family: the words a field uses, what each means here, how it differs from its neighbours, and what people say instead. The line grammar, the sense model, where it feeds, and how to build one.
• Semantic indexing: two legs and a fusion — A docset can be asked a question by token or by meaning, and the two ways fail in opposite directions. What each leg costs, where each breaks, why reciprocal-rank fusion is the default, and how to read the recorded run at /demo/.