@@ -113,7 +113,7 @@ Tokenizer/context fit, RAM/VRAM, provider access and all live limits need later
 `npm run simulate:commons -- --seed 17` exercises actual disposable HTTP transactions,
 stale writes, competing named fixture actors, rejected review, timeout, cancellation,
 provider failure, invalid output, pagination and disk restart recovery. Default
-history 12 yields 48 events across 16 pages; history 40 yields 80 events across 27 pages. Actors and
+history 8 yields 48 events across 16 pages; history 40 yields 80 events across 27 pages. Actors and
 outcomes are explicitly simulated. [Commands and evidence boundaries](SIMULATIONS.md).
 
 ## Sandbox and tool boundaries
 
@@ -159,4 +159,6 @@ Final integration only: consolidate handoff/context/launch drafts, freeze all so
 - Final integration began before the reserved window. The handoff was condensed into an operational summary while retaining exact implementation IDs, architecture entry points, activation requirements, reproduction commands, data boundaries, failures, source publication workflow and a continuation prompt. Context stays at 119 doctor lines and now records the user-opened app and current evidence lookup instead of the stale Timeline commit. Launch drafts retain channel rules and gain a factual technical addendum. No new feature or model integration is started.
 - A read-only 14:02 retained-state check confirmed scenario revision10/step10, Commons55,9 jobs/95 events and the same approved-unmerged Safe Timer PR. Model/identity/backups retained earlier measured sizes; the database directory grew 73,728 bytes after the explicit normal-app startup, so no unchanged-byte claim is made. Final Docker doctor still cannot reach a daemon and qualifies no container or isolation. Linux CI receipts for public `837a3d1` were downloaded and matched the clean archive's normalized hash `1312ad7b8e0ddb7adaea299cd15c5a5df4addffa4ff790089746b251a12c70b8`; both differ from the maintainer's actual Windows working bytes as documented.
 
-Freeze all source/docs/assets for the final delivery full gate and separate 56-case browser suite. Preserve immutable delivery receipts, commit documentation, verify the post-commit v2 hash, publish only through the checked separate-history workflow and inspect exact CI. Leave the user's normal app/tab available. No further feature work; deadline 15:46:17 UTC.
+- The first documentation delivery passed 108/2 DOM/56 Chrome on source hash `a75372cef0fb9d042b19cb71f5aefcd5d381057d58839f85ea734e2fc98fe9e5` and published local `0a0d9488a1ddf16165e1e066c8c540262600e9a5` / public `5540dbf6b85d4b045b98dfec55228165342aab45`, equal tree `4faef7e7302f6a8a9c21a9073c099ea5461147b9`. CI39268489148 passed all three jobs, and actual full/browser artifacts plus fresh GitHub preflight were retained. During final manifest cross-check, the handoff's default simulation history was found to say12 instead of the actual8 in both CLI/function and stored seed17 report. Correcting that factual documentation error preserves the actual48 events/16 pages; history40 remains80/27. No application code changes. Retain the preceding evidence and repeat the mandatory frozen full/browser gates under distinct `delivery-final-*` receipts for the corrected source.
+
+Freeze corrected documentation and all other source/assets for the final full gate and separate 56-case browser suite. Commit only this factual repair after acceptance, verify the post-commit v2 hash, publish through the checked separate-history workflow and inspect exact CI. Leave the user's normal app/tab available. No further feature work; deadline 15:46:17 UTC.