title: UCCA-CORRECTION — the proven run's identity, corrected from bytes doc_id: UCCA-CORRECTION-01 status: READY to cross (accompanies the §4.1 schema sendback). Corrects one factual claim that propagated through the crossed record; per FENCE-PROTOCOL-01 §3, correction is by new crossing content — no crossed original is edited. layer: crossing artefact (UCCA original) authored: 2026-07-01 NYC — engine side; ruled by Tim (testimony: no memory of any Dec-2025 retail run); byte-ground by Alex (engine repo walk, no such bundle exists) relates_to: UCCA-FOUNDATION-01 §4 (origin of the claim; corrected same date); UCCA-ASSESSMENT-RTOPACKS-BRIEF-01 R-1; RTOPACKS-ENGINE-BRIEF-01 R-1 (carried the claim in good faith); UCCA-CONTRACT-PACK-01; RTOPACKS-CORRECTION-01 (the pattern this mirrors)
Correction: the proven generation run¶
C-1. The false claim and its provenance¶
"Generation proven on a retail unit, Dec 2025" is false as an on-disk claim. No such run bundle exists anywhere on the engine repo. The claim originated in UCCA-FOUNDATION-01 §4 (authored from session memory), was carried into the engine's assessment (R-1), your brief (R-1), and the contract pack — four crossings, unverified at each. Caught engine-side during the §4.1 schema extraction; ruled by Tim 2026-07-01.
The fuller ruling (founder testimony, filed as such): a retail test run did occur around Dec 2025 — JSON output only, nothing persisted to a database — and its files were most likely deleted (or unlinked) in the workspace tidy-up when the project moved. The historical event is attested; the artefact is gone. Under the honest ledger, an attested-but-unhashable run is history, not proof — it may not be cited as validation. A short engine-side search for unlinked output is queued; recovery would be noted but changes nothing below.
C-2. The true run (byte-confirmed)¶
Generation is proven — on CHCPRT025 (a care/community-services unit), run 2026-03-09, the validated, provenance-tracked bundle on disk from which the §4.1 schema accompanying this crossing was extracted. Every load-bearing statement built on the proof stands unchanged: generation works end-to-end on one real unit; the run bundle's provenance shape is the §4.1 raw material; R-1's YES is unaffected. Only the unit's identity and date were wrong.
C-3. Housekeeping both sides¶
Engine-side: FOUNDATION-01 §4 corrected with a dated changelog line; fossil-ledger row closed (how-we-know: bytes + testimony). Your side: wherever your records echo "retail unit, Dec 2025," read "CHCPRT025, March 2026" — at your convenience; nothing built depends on it.
Third false claim retired today, and the most instructive: it lived in the document written to be true, passed a byte-audit that didn't cover it, and crossed the fence four times on canon's authority. The run it described probably happened — and that is exactly the lesson: an event without an artefact is a memory, and the ledger deals in artefacts. The citable proof is CHCPRT025; the Dec-2025 run is filed as attested history. The record is now true.