jeffryandClaude Opus 5
f026f3ea55
Someone in an empty repository saying "set this up" has nothing to read. The four primitives say what things are and deliberately not what to do, which leaves a set of primitives with no starting point — a puzzle rather than a foundation. Two questions carry it: what can a reader not already know, which is what you publish; and what do you need that nothing provides, which is a gap. Everything else is conditional on having something to say, and the page says plainly that a file carrying no delta should not exist. It ends by pointing at jeffry/homelab-cluster and saying that if the two disagree, the example is right. Precedent beats description, and a page about writing documents should not be the longest document. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
bedrock
The primitives. Four documents, in .loom/published/:
- The
.loomdirectory — everything lives there, and a directory means a kind - Sibling facets —
x.<facet>.mdbesidex.md, owned by whoever wrote the facet - Recording decisions — what an entry says, and why it is never revised
- Publication — what
.loom/published/means - Starting one of these in a repository — the one page here that tells you what to do, because a set of primitives with no starting point is a puzzle
Nothing else here says what to do. loom/annotating, loom/cart and
loom/externals are built on these and are where the practices live.
Most repositories will never read this. Its audience is people building conventions, not people using them — and it is a good sign that a primitive is one nobody needs.
Publishers are the exception, because publication is defined here, and the people whose choices land on somebody else should have read the primitives.