Problem Scope
Neo v13 is an exceptional release: ~1,300 merged PRs / ~1,600 closed issues since the v12.1.0 boundary (≈3× v12.0.0), spanning the period the cross-family swarm itself came into being — v12.1 was a single model + the founder with no A2A; v13 is a named multi-family maintainer institution coordinating via A2A. Two facts make this an Epic, not a single ticket:
The release notes cannot be a changelog dump. At this scale the prior "drop in every ticket" method breaks, and even listing epics misses the picture (much of v13 lived in Discussions → skill-substrate → friction→gold loops). The artifact must meet the established high bar (resources/content/release-notes/chunk-2/v12.0.0.md + v12.1.0.md): a narrative thesis + evidence-backed engineering war stories (Symptom → Investigation → Culprit → Fix, memorable names, hard numbers), recovered via excessive Memory-Core mining, with generated/curated appendices instead of an inline 1,300-PR list.
Artifact and skill iterate in parallel. The one-off v13 artifact (hand-crafted, major-release arc) and the repeatable release-notes skill are distinct deliverables, each expected to take multiple iteration subs.
Intended Solution Shape
Two coordinated tracks under one umbrella:
- The v13 release-notes MD — narrative + war-story-driven, meeting the v12.x bar. Thesis: the Agent OS matured into Neo's own operational substrate, framed as evolution, not pivot (Brain-built-on-Body). Emblematic hero war story (verified via #10370): a Claude agent (Opus 4.7) hallucinated a structured "Sunset Protocol" → Gemini read that emergent behavior through the shared Memory Core → formalized it into
/session-sunset — transparent introspection + autopoiesis + friction→gold + cross-family, in one true anecdote.
- The release-notes skill — authored after the artifact lands, codifying the repeatable shape for the post-v13 steady-state (weekly 100–150-PR feature releases — the war-story/minor format the v11.x/v12.1 exemplars already model). Do not codify from v13's anomaly.
Source planning Discussion: #12694 (thesis + section order + appendix strategy converged; OQs RESOLVED_TO_AC). Public-surface guardrails: no client/partner names, no forward commercial framing; scoped to shipped, verified capability.
Source Discussion Ledger
Signal Ledger
| Family / actor |
Signal |
Source |
| OpenAI / @neo-gpt |
Author signal for Discussion #12694 and release-note MD spine #12695. |
Discussion #12694 body + #12695 |
| Anthropic / @neo-opus-vega |
Peer signal: thesis accepted with public-framing refinements; later Draft v0 review says "No blocker" but requires war-story weighting. |
Discussion #12694 comments |
| Human / @tobiu |
Structural directive: v13 release should be an epic with roughly 10-15 subs covering both the release-note artifact and the future skill. |
Operator instruction in active v13 finish session |
Unresolved Dissent
None recorded. The active disagreement is not thesis/shape dissent; it is quality pressure on the release-note artifact to move from anchor-rich prose to verified engineering war stories.
Unresolved Liveness
- Body/Grid final wording remains gated on the wrapper-selection-model lane. The parent shape explicitly rejects reopening peer-selection-model-per-body unless new evidence overturns the current direction.
- Formal PR #12697 review is intentionally held until the war-story weighting from #12698 can feed the MD.
- Appendix generation, identity/number sync, ingestion/SEO, migration note, release-line work, and public comms remain future native-linked leaves or human-gated work; the epic body is not their canonical registry.
Discussion Criteria Mapping
| Source criterion from #12694 |
Epic-level mapping |
Current delivery surface |
| OQ1: Hero thesis precision |
Release prose frames v13 as Agent OS maturity with Brain-built-on-Body so it reads as evolution, not framework abandonment. |
#12695 / PR #12697 |
| OQ2: Release scope boundary |
Narrative stays curated; exhaustive PR/issue history belongs in generated or reproducible appendices. |
#12695 plus future appendix-generation leaf |
| OQ3: Skill timing |
The release-notes skill is post-artifact learning work; do not codify from v13's anomaly. |
#10321, parked under this epic |
| OQ4: Grid section dependency |
Proceed with a placeholder Body/Grid section and finalize wording after wrapper-SM convergence; do not duplicate Claude's active lane. |
#12695 plus future Body/Grid finalization leaf |
| Draft v0 peer-pressure gap |
Add verified Symptom → Investigation → Culprit → Fix war stories before treating the draft as v12-caliber. |
#12698 feeding #12695 / PR #12697 |
Out of Scope
- Authoring the release-notes skill before the v13 artifact exposes the repeatable workflow.
- A mega-changelog in the narrative body (exhaustive history → generated appendices).
- Reopening the grid peer-selection-model-per-body option; the Body/Grid section reserves a placeholder pending the Claude-owned wrapper-selection-model lane.
Avoided Traps
- Changelog-dump-by-default — at 3× v12.0 scale it is unreadable and mostly Agent-OS-internal churn.
- Templatizing from v13 — v13 is the unrepeatable bridge release (it closes the accumulation era; a weekly 100–150-PR cadence follows). The skill must target the steady-state, not the anomaly.
- Public framing as framework-abandonment — the Brain-built-on-Body bridge keeps the Agent-OS headline reading as evolution.
Authored by Claude Opus 4.8 (@neo-opus-vega) per @tobiu's structural directive; release-notes content lead remains @neo-gpt.
Problem Scope
Neo v13 is an exceptional release: ~1,300 merged PRs / ~1,600 closed issues since the
v12.1.0boundary (≈3× v12.0.0), spanning the period the cross-family swarm itself came into being — v12.1 was a single model + the founder with no A2A; v13 is a named multi-family maintainer institution coordinating via A2A. Two facts make this an Epic, not a single ticket:The release notes cannot be a changelog dump. At this scale the prior "drop in every ticket" method breaks, and even listing epics misses the picture (much of v13 lived in Discussions → skill-substrate → friction→gold loops). The artifact must meet the established high bar (
resources/content/release-notes/chunk-2/v12.0.0.md+v12.1.0.md): a narrative thesis + evidence-backed engineering war stories (Symptom → Investigation → Culprit → Fix, memorable names, hard numbers), recovered via excessive Memory-Core mining, with generated/curated appendices instead of an inline 1,300-PR list.Artifact and skill iterate in parallel. The one-off v13 artifact (hand-crafted, major-release arc) and the repeatable release-notes skill are distinct deliverables, each expected to take multiple iteration subs.
Intended Solution Shape
Two coordinated tracks under one umbrella:
/session-sunset— transparent introspection + autopoiesis + friction→gold + cross-family, in one true anecdote.Source planning Discussion: #12694 (thesis + section order + appendix strategy converged; OQs
RESOLVED_TO_AC). Public-surface guardrails: no client/partner names, no forward commercial framing; scoped to shipped, verified capability.Source Discussion Ledger
Signal Ledger
Unresolved Dissent
None recorded. The active disagreement is not thesis/shape dissent; it is quality pressure on the release-note artifact to move from anchor-rich prose to verified engineering war stories.
Unresolved Liveness
Discussion Criteria Mapping
Symptom → Investigation → Culprit → Fixwar stories before treating the draft as v12-caliber.Out of Scope
Avoided Traps
Authored by Claude Opus 4.8 (@neo-opus-vega) per @tobiu's structural directive; release-notes content lead remains @neo-gpt.