Merged PR census: 3 / 15 / 73 / 92 across October 1–7 / 8–14 / 15–21 / 22–31. The count includes mirror, CI, tests and documentation—not 183 user-facing features or a measure of effort. Final CI jobs means Validate, deployed-stage E2E and Deploy Production in one exact build. No monthly line-churn total is asserted.
The final October merge, ceffbf327, completed CI at 19:45:36 UTC on October 31. Validate, deployed-stage E2E and Deploy Production all succeeded. This supersedes the earlier pending #397 snapshot and the failed intermediate #369 run.
It proves those jobs ran successfully—not independent production browser acceptance, package installation, or owner approval.
GitHub release records were observed for 0.2.0 and 0.2.1 on October 14, 0.2.2 on October 26 and 0.2.3 on October 29. Seven packages have 0.2.3 records.
Later merged conversion, upgrade and preview work must not be described as a verified new package release. No independent npm install was performed for this report.
The repository evidence establishes specific builds and test outcomes, not completion of the entire connect → edit → save → preview → publish journey across all supported environments. The remaining open PRs and package rollout conditions must be assessed separately.
#362 remained open with a new October 31 head: validation passed, but deployed-stage E2E failed. #363 remained an open draft despite successful exact-head CI. Neither is counted as a merged October result.
Mirror workflow and test configuration were repaired; preview-session diagnostics were added without claiming a behavior fix. The separate landing integration supplied CMS blocks, routes and shared site data for a more representative editing surface.
Billing-off removed an unmet production dependency while retaining billing coverage in staging. Public cloud-astro packaging expanded what users could install. Two regression batches then repaired rich text, fragments, indexing and related behavior found in 0.2.0; #210’s validation and deployed-stage E2E were independently checked.
The focus moved to writes: tab isolation, safe Git commits, non-overwriting media moves and correct MDX write-back. Refresh and workspace security received targeted fixes. Generated Typia contracts and repository commands through rig.yaml improved consistency, but did not finish the separate StateBus migration.
Invite routing, page state, block actions and undo gained corrections and regression coverage. GitHub authority was reconciled with existing workspaces, not just fresh previews. The How it works site update landed after browser proof; dependency-review and mirror workflows moved to a self-hosted pool.
Keyless editing was followed by sharing, static-site previews, agent tooling and consistent media processing. GitHub reinstall recovery and large-push handling addressed operational failures. After 15:24 UTC, #398 added queued imports, #400 added a doctor warning and #401 fixed a DOM-test teardown race. The final default-branch CI and production job succeeded before the month closed.
Evidence is classified by what was actually read. Historical PR proof is useful, but it is not a new run of the product. No review-round counts or approval claims have been reconstructed from incomplete records.
| Area | Observed proof | Boundary / next check |
|---|---|---|
| Write isolation | PR #224: exact-head Validate and deployed-stage E2E passed. CI evidence. | Confirms that tested snapshot, not every later build or all multi-tab environments. |
| Content parsing | PR #239 reports red/green regressions; successful Validate and deployed-stage E2E were checked independently. CI evidence. | This verifies a specific hardening change, not closure of the entire security backlog. |
| Keyless editor | PR #344 reports Chromium/Firefox/WebKit, static-output parity and draft-preview proof; exact-head validation/E2E passed. CI evidence. | Live Vercel was not exercised; reported cold-open time was 6–7 seconds against a three-second target. |
| Docs/gallery #362 | Head aa056647, committed October 31 at 21:46 UTC: Validate succeeded, deployed-stage E2E failed and production was skipped. The log shows an editor-page assertion timing out. Failed E2E job. | Still open. The earlier lint/artifact failure is historical, not the current blocker. A fresh stage was deployed in this run; its current browser availability was not tested. |
| Publish summary #363 | Head 76bb09e: Validate and deployed-stage E2E succeeded; production skipped. Exact-head CI. PR reports 52 Nx tasks, 226 Storybook tests and 162 real-Git-runtime tests. | Still an open draft. Preview expired October 30; authenticated publication and team acceptance remain unestablished. |
| Final pipeline | ceffbf327: Validate, deployed-stage E2E and production deployment all succeeded. Final October run. | Do not infer owner acceptance, customer-site upgrades or new public package publication from CI success. |
The next coordinated package release and real-site upgrade checks remain necessary. Repository quotas were implemented without enabling a quota value in that PR. The affected repository’s recovery, preview-site updates and sender-domain readiness require operational confirmation.
StateBus #84 remained an open draft in the collected evidence; full-page docs preview is still unavailable. Evidence retention also remains a separate issue: a path in a log or a green job is not proof that screenshots are retrievable now.
The expandable rows below are a deliberately small, verified sample: two files from #197, one from #398 and one from #400. Counts are per-PR diffs, not month-wide unique-file totals. Links point to inspected lines at the final October default-branch commit. This avoids inventing a monthly diff from a PR census.
A navigable selection of real merge commits behind this narrative. Captions describe the associated PRs; this is not a complete commit log.
The following records summarize decisions evidenced by merged PRs. They are report summaries—not newly ratified project ADRs.
The verified merged-PR author counts are 89 for alex-mextner, 80 for NikoT2 and 14 for other authors. These are repository authorship counts—not sole-credit, personal activity or effort measurements. PR census.
The separate October 6 landing/CMS integration and How it works website update are credited through their repository records rather than private correspondence. No customer-acquisition, revenue or campaign result is asserted.
Recommendations from the evidence gaps—not new commitments assigned to the team.
Source context: PR #362 · PR #363 · PR #377 · PR #395 · PR #320.
Reporting window: October 1–31, 2026 (UTC). The original repository evidence ended at October 31, 15:24 UTC. A targeted November 1 read-only refresh covered all October 31 merged PR search results, selected exact-head/default-branch workflows and the latest public GitHub releases. It added three merges and replaced stale pending/failed-build descriptions with their terminal outcomes.
Repository evidence includes the month’s PR census, 66 directly checked PR records in the original evidence set, selected jobs, public release records and the landing commit. This report is a selective technical synthesis, not an exhaustive audit of every branch, ticket, deployment or user scenario. No new product tests, npm installation, production browser audit, business-metric estimate or screenshot recapture was performed.
Some repository evidence links are access-controlled and require permission to view. This public report contains no private correspondence, customer identifiers, credentials or detailed security exploit mechanisms. Merge, release, deployment and product acceptance remain distinct evidence levels.