field notes KEEL

the repository is not the fleet

The declarative path existed and lagged the imperative one, in the ordinary way: a change typed by hand at the moment of need, a repository updated later, or not at all. Most changes were still typed by hand, and the written record said which ones.

The evidence lives in the diffs. One script in the repository differed from the copy that had been running for months, in ways that mattered, and the instruction recorded for that situation was to diff first, prefer the live copy, and commit the fix — the running artifact is the source of truth until it is proven otherwise. Elsewhere a comment in the repository confidently described a layout that had been gone for weeks, and the correction happened only because someone went looking for it during an unrelated cleanup.

The counterweight is measurement. A sweep compared every running definition against the repository and found no drift, which is only a finding if the comparison is possible and repeatable. Floating references and stale definitions were flagged in the same pass, which is what turned a green tick into a list.

What this proves is that a repository is a claim about the fleet that has to be renewed, not a description that stays true by itself. The honest version of "partly automated" is a list of which parts; the dishonest version is a plan that describes a migration as though it had already happened.

drawn from: repository-versus-live audit commits

Every note carries a codename and one idea. What the systems implement is withheld by design — see the index for the rules.