Skip to main content

Publish

Make it available — then go and look. The result is put where the persona can reach it, and its arrival is then confirmed by querying for it.

Completion is not reported, it is found. A report that something shipped is itself a statement with no evidence behind it, and the practice does not accept one anywhere else either. So the last act is to look: is the thing actually there?

The capability is produced here, and produced automatically — generated from the specifications that passed, never authored by hand. What the system is said to do therefore cannot drift from what was proven about it, because nobody is in a position to write the first without the second.

Complete is a state, not an act. An item is complete because its publication was confirmed — not because anyone marked it so. Nothing in the practice declares its own completion.

What Complete does not mean​

It does not mean anybody has it. In much of the industry generally available means customers have it. Read that way, this state reproduces exactly the failure the practice exists to end — a claim that somebody has something, resting on the fact that it was released. The closer analogue is released to manufacturing: it exists, it is retrievable, and taking it is somebody else's act.

It does not mean the work was right. An item at Complete may still turn out to be wrong, and finding that out raises a fix attached to the capability it stated. Nothing is reopened: the track runs once per item, and a correction is its own item travelling it again.

If publishing routinely runs ahead of verification

Verification becomes a formality nobody blocks on. What is published-but-unverified has to be distinguishable by whoever consumes it — a prerelease, a separate channel, a tag. Without that distinction the gate has no teeth, and what the system is said to do fills with things nothing confirmed.

Done when: the published thing has been found by looking for it.