#19712 is the epic for Engine 13.3.0's release notes; iteration 3 merged as #19760 (e127a36d63). On 2026-10-15 the operator relayed outside feedback: a technical reader new to Neo found the 13.2 notes hard to follow, although their content impressed. The operator's read is that the team "might be too much 'lost in the woods' or biased, since the team created the code". Three independent reads of iteration 3 followed on #19712, each as a CS-trained engineering lead with no Neo context: Julia's, Emmy's, and a context-free reader's, given only the two public notes. They converge.
The Problem
All three reads name the same failures, with line anchors in each comment:
The note never says what Neo.mjs is, nor why content-sized rows are hard in it: the grid's logic runs in a worker without a DOM, so it cannot measure text.
The TL;DR opens on 13.2's statistics and bills the release process beside the two grid features.
Build durations, line counts, CI run IDs and review rounds recur through the body. An outside reader takes them for haste.
Insider terms go unexplained: start, center and end bodies; pinning origin; Brain-gated; host leaf; executed discriminator; refusal codes; hydration; Group projection, participant and document; fold; "the operator"; D#N.
The release-cut and CI chapters are written for maintainers, yet sit among the hero chapters.
The adopter's first question goes unanswered: who builds this, and what keeps AI-written code correct.
The adopter limits share a list with a CI retry, and the Chromium 156 grid fix, an upgrade argument, sits in Also.
The narrative ends on an upgrade table, and its "open" links lead to GitHub source rather than a running example.
The Architectural Reality
The note is .github/RELEASE_NOTES/v13.3.0.md, iterated on dev behind its banner (release-notes §1, §6). The designation lines stay as the operator confirmed them (§4). Generate.spec.mjs refuses blob/main and tag-ref source links.
neomjs.com builds from the published package: examples/grid/bigData answers there today, while examples/grid/issues and examples/grid/contentSizedRows answer once 13.3.0 publishes, the first provided #19761 merges before the cut.
The README's maintainer table maps names to handles and model families and states the cross-family review. In this window the operator merged every pull request except four Dependabot bumps, which merged automatically (live GitHub, 2026-10-15).
ROADMAP.md's 13.3 section lists what is deferred; 13.4's scope is not set.
The Fix
Rewrite for a technical reader new to Neo, keeping every engineering story, source and bound:
Orientation first: a TL;DR from the reader's problem (clip the text or pad every row) to the two opt-in features; a short New to Neo.mjs? (the AI team and the engine it builds; app logic in a worker, the DOM on the main thread) that says who builds it: names mapped to handles and model families, cross-family review, the human merge.
Chapters around guarantees: content-sized rows (why measuring is hard, measured by the browser, keeping your place, the one-pixel font case, the recycled-row case, bounds); header menus (counts under the other filters, independent filter ownership, typed values, bounds); a reliability chapter that opens on a green suite driving the wrong app and keeps the async-predicate war story, retitled as the general pitfall.
Process out of the arc: build durations, line counts, CI run IDs and review attributions leave the body. The release-through-PR chapter becomes a section labeled for maintainers, after the main arc. The changelog and credits keep the source trail.
Each term listed above is glossed on first use or replaced by a plain name.
The Chromium 156 fix moves into Upgrading.
Ending: a runnable journey through the showcase on neomjs.com, then a forward view from the roadmap's deferred list that promises no 13.4 scope.
Changelog: the merges since iteration 3, counted against live GitHub at a stated time.
Acceptance Criteria
AC-1 Before the first chapter, the note says what Neo.mjs is, who builds it, and why content-sized rows are hard in it; names map to handles, and each claim about review and merging cites its source.
AC-2 The TL;DR and 13.3 in 2 Minutes lead with the reader's problem and the two opt-in features; the release process takes at most one line there.
AC-3 Above Full Changelog, the body carries no build durations, line counts or CI run IDs; one scale line remains in 13.3 in 2 Minutes.
AC-4 Each insider term listed in The Problem is glossed on first use or replaced.
AC-5 The release-through-PR material sits in a section labeled for maintainers, after the main arc, and keeps its unperformed-cut bound.
AC-6 The narrative ends with a runnable journey through examples/grid/issues and a forward view that promises no 13.4 scope; the adopter limits are stated, and the Chromium 156 fix is in Upgrading.
AC-7 The Full Changelog lists every pull request merged since iteration 3, counted at a stated time.
AC-8 The five unit specs that read release notes pass: Generate.spec.mjs, PublishReleaseNoteOrphan.spec.mjs, RebuildContentIndexesAndSeo.spec.mjs, WorkflowScopeClassifier.spec.mjs and developmentThemeAssets.spec.mjs.
AC-9 A GPT-family review approves before merge (release-notes §5).
Out of Scope
The cut's receipt, removing the banner, and the numbers at the cut head: #19765, under #19652.
The showcase and its data: #19756 / #19761 and neomjs/pages#44.
The release-notes skill's bar for velocity numbers and outside reads: a ticket of its own in neo-agent-skills, if the team agrees.
Avoided Traps
Writing down to the reader. The reader is technical; the depth stays, and the context comes first.
Cutting the war stories. They are the most portable lessons; they move to principle-first framing.
A roadmap the team has not set. The forward view names deferred work as deferred.
Peer reads as the outside view. Every peer is an insider; the context-free read is the control.
Sweeps at 2026-10-15 17:18Z: the latest 20 open issues hold no iteration-4 leaf; #19765 (cut-prep) and #19766 (the release PR) start after this one merges. A2A, latest 30 in all states, holds no competing claim. Memory Core ("13.3 release notes hard to follow for outside readers, too long, jargon, durations read as rushed") returns the narrative reads and earlier notes feedback, and no decision against this rewrite. My open assignments on this surface are #19712 and #19756; neither overlaps.
Context
#19712 is the epic for Engine 13.3.0's release notes; iteration 3 merged as #19760 (
e127a36d63). On 2026-10-15 the operator relayed outside feedback: a technical reader new to Neo found the 13.2 notes hard to follow, although their content impressed. The operator's read is that the team "might be too much 'lost in the woods' or biased, since the team created the code". Three independent reads of iteration 3 followed on #19712, each as a CS-trained engineering lead with no Neo context: Julia's, Emmy's, and a context-free reader's, given only the two public notes. They converge.The Problem
All three reads name the same failures, with line anchors in each comment:
D#N.The Architectural Reality
.github/RELEASE_NOTES/v13.3.0.md, iterated ondevbehind its banner (release-notes §1, §6). The designation lines stay as the operator confirmed them (§4).Generate.spec.mjsrefusesblob/mainand tag-ref source links.examples/grid/bigDataanswers there today, whileexamples/grid/issuesandexamples/grid/contentSizedRowsanswer once 13.3.0 publishes, the first provided #19761 merges before the cut.ROADMAP.md's 13.3 section lists what is deferred; 13.4's scope is not set.The Fix
Rewrite for a technical reader new to Neo, keeping every engineering story, source and bound:
Acceptance Criteria
examples/grid/issuesand a forward view that promises no 13.4 scope; the adopter limits are stated, and the Chromium 156 fix is in Upgrading.Generate.spec.mjs,PublishReleaseNoteOrphan.spec.mjs,RebuildContentIndexesAndSeo.spec.mjs,WorkflowScopeClassifier.spec.mjsanddevelopmentThemeAssets.spec.mjs.Out of Scope
Avoided Traps
Related
Parent: #19712 · iteration 3: #19759 / #19760 · the showcase: #19756 / #19761 · the cut: #19652
Sweeps at 2026-10-15 17:18Z: the latest 20 open issues hold no iteration-4 leaf; #19765 (cut-prep) and #19766 (the release PR) start after this one merges. A2A, latest 30 in all states, holds no competing claim. Memory Core ("13.3 release notes hard to follow for outside readers, too long, jargon, durations read as rushed") returns the narrative reads and earlier notes feedback, and no decision against this rewrite. My open assignments on this surface are #19712 and #19756; neither overlaps.
Origin Session ID: 8779ddeb-f661-411b-b93d-c3b6c2158b11
Retrieval Hint: "v13.3.0 release notes iteration 4 outside reader", "release notes insider vocabulary", "what Neo.mjs is in the release notes"