Shipped live

The intent/1 specification with its JSON Schema; the reference library and CLI (adopt · file · promote · close · emit · merge · verify · registry, zero runtime dependencies); the reference consumer — the spec's Discovery rules as code, so "unreachable is not absent" is a function, not a paragraph; the registry mechanism with its fetch-check in CI; the dependency-free vendored checker; the legacy backlog/1 compatibility promise; and this site, which makes zero external requests, like the format it describes.

Gate 1 — the first independent adopter open

An organisation with no financial relationship to the steward's group serves a valid fragment at its well-known path. This gate is what the registry counts, and it is currently zero — printed there at its true value. Everything below waits on it, because building a standard's amenities before anyone independent uses the standard is how formats die with excellent documentation.

Gated on adoption

A public merge view — the pieces exist (fetchFragment over registry.json into merge) and the roll-up ships the day the registry has members, not before: a published projection over nothing teaches every reader the format is decorative. A conformance badge — granted by a fetch, never typed by hand. A governance body with outside seats — promised in governance for the moment production adopters want a say, and deliberately not before. intent/2 — if and when real usage demands a shape change, with a reader that accepts intent/1 for as long as anyone emits it.

The kill criterion, in public

If no independent organisation is publishing within a real pilot window, the format stays excellent internal infrastructure and no further standard-building is funded. A standard that cannot find one stranger is not early — it is answered. Writing that here is the point: a roadmap that cannot fail is a brochure.