Version. 0.1 Date. 2026-08-20 Status. Standing vocabulary correction. Applies to all engine-side work, records and reports from CR-2026-230 onward. Read before writing any CR, close report, or acceptance criterion. Origin. Operator ruling, 2026-08-20, at the W2 close of the H0 lane.
Drop the word "demo" from engine-side work, records and reports.
The word was covering two different things, and they separate cleanly.
The seven-step traversal — mint, work, warn, stop, freeze, unfreeze, walk the record — is an end-to-end integration test of H0. It is real work and it keeps its acceptance role. Call it the H0 end-to-end traversal.
Presentation artifacts are Boundaries-side communication work. They are not engine work, and they do not sequence engine decisions.
A word that means both "the test that proves the system works" and "the thing we show people" lets the second quietly set the priorities of the first. Once that happens, every engineering question acquires a second, unstated criterion — how will this look — and the two criteria do not always point the same way.
The clearest case, from this lane's own record. The traversal's work step runs a shaping job rather than a rendering one. Under the old vocabulary that reads as a choice about what to show, and invites the suspicion that the safer path was picked. It is not that. Shaping is where the identity model is complete; rendering's is not (per-binding identity versus per-role sessions — see loomworks-scoping-per-binding-session-minting-v0_1.md). The traversal exercises the path where the model is finished, which is the only thing an integration test can honestly exercise. The old word made a true engineering statement look like a presentational one.
"An unlabeled agent action is impossible" is true or false of the runtime. It is not a claim about a script, a rehearsal, or an occasion. Nothing about it is satisfied by arranging for the false case not to arise while someone is watching.
Consequences for how criteria are written and reported:
2026-09-09 is the date defended for H0 complete and traversable end to end. It is a forcing function, not an event.
Nothing happens on that date that the system does not already do on 2026-09-08. The date exists to make scope decisions expensive to defer, which is what a forcing function is for. Reporting against it means reporting what the runtime does, not what could be arranged to happen once.
Do not rewrite closed CRs. CR-2026-226, CR-2026-227 (v0_1–v0_6) and CR-2026-229 use the old word throughout, and they stay as written — the same never-overwrite discipline that keeps superseded versions as siblings.
This note is where a future reader meets the correction. A reader arriving at "the demo script" in a closed CR should be able to find this and know that the phrase means the H0 end-to-end traversal, and that the sequencing decisions recorded there were engineering decisions.
The prior framing is preserved rather than smoothed for the usual reason: the correction is more informative than the corrected text would have been. That the word needed separating — and that it was doing work nobody had noticed — is itself a finding, and erasing the evidence would leave only the conclusion.
| Instead of | Write | |---|---| | the demo script | the H0 end-to-end traversal | | demo assembly | traversal assembly, or the remaining H0 work | | before the demo | before H0 is complete | | the demo date | the date H0 is complete and traversable | | what the demo shows | what the runtime does | | a risk for the demo | a property the runtime does not yet have |
Related: loomworks-standing-note-descriptions-of-state-are-not-state-v0_2 — the same family. There, a description of state was mistaken for state; here, an occasion was mistaken for a property.