Part of #14310. Refs #14313 (the merged front-door guide).
Problem
§5's headline proof — "more than 1,000 merged pull requests in a single month" — invites a fair skeptic: can one human gardener gate ~33 PRs/day without rubber-stamping? The objection is valid only if the human reviews are full-diff reviews. They aren't. Every PR is cross-model peer-reviewed (often several rounds to approval), and passes a battery of CI pipelines + pre-commit hooks, before it reaches the human — whose gate is strategic (right thing to build? matches Neo's architecture + best practices?), not line-safety. §5 already states cross-model-review-before-human, but not the CI / pre-commit / multi-round guardrails that make the volume tractable — leaving the skeptic's math open.
Surfaced by an external Gemini Flash 3.5 review of the merged guide (no Agent OS connection — a cold read). V-B-A'd against the repo: the guardrails are real — 20 workflow files / 21 jobs in .github/workflows/ (~14 PR-quality gates: unit/integration test.yml, CodeQL, plus a dozen lint/structural guards); husky pre-commit + pre-push + lint-staged confirmed.
Fix
Augment the §5 bullet to fold in the guardrails (multi-round cross-model review, more-than-a-dozen CI pipelines, pre-commit hooks) and make the gardener's gate explicitly strategic — without dropping the unattended-overnight liveness beat (the external review proposed swapping it out for a CI enumeration; rejected — "runs the night shift with no operator awake" is the section's strongest card, and every shop has CI). Conceptual framing ("more than a dozen CI pipelines"), not a brittle hard count that dates.
Acceptance Criteria
Files
learn/benefits/WhatIsNeo.md (body only — no tree.json).
Coordination
The IA reposition (Introduction.md → ApplicationEngine.md + nav) is #14311 (@neo-gpt, blocked on #14312). This ticket deliberately does not touch tree.json, so #14311 rebases cleanly over the new §5 text — no collision. A2A sent.
Origin: external Gemini review + @tobiu direction.
Part of #14310. Refs #14313 (the merged front-door guide).
Problem
§5's headline proof — "more than 1,000 merged pull requests in a single month" — invites a fair skeptic: can one human gardener gate ~33 PRs/day without rubber-stamping? The objection is valid only if the human reviews are full-diff reviews. They aren't. Every PR is cross-model peer-reviewed (often several rounds to approval), and passes a battery of CI pipelines + pre-commit hooks, before it reaches the human — whose gate is strategic (right thing to build? matches Neo's architecture + best practices?), not line-safety. §5 already states cross-model-review-before-human, but not the CI / pre-commit / multi-round guardrails that make the volume tractable — leaving the skeptic's math open.
Surfaced by an external Gemini Flash 3.5 review of the merged guide (no Agent OS connection — a cold read). V-B-A'd against the repo: the guardrails are real — 20 workflow files / 21 jobs in
.github/workflows/(~14 PR-quality gates: unit/integrationtest.yml, CodeQL, plus a dozen lint/structural guards); husky pre-commit + pre-push + lint-staged confirmed.Fix
Augment the §5 bullet to fold in the guardrails (multi-round cross-model review, more-than-a-dozen CI pipelines, pre-commit hooks) and make the gardener's gate explicitly strategic — without dropping the unattended-overnight liveness beat (the external review proposed swapping it out for a CI enumeration; rejected — "runs the night shift with no operator awake" is the section's strongest card, and every shop has CI). Conceptual framing ("more than a dozen CI pipelines"), not a brittle hard count that dates.
Acceptance Criteria
ai:lint-guides learn/benefits/WhatIsNeo.mdgreen.Files
learn/benefits/WhatIsNeo.md(body only — notree.json).Coordination
The IA reposition (
Introduction.md→ApplicationEngine.md+ nav) is #14311 (@neo-gpt, blocked on #14312). This ticket deliberately does not touchtree.json, so #14311 rebases cleanly over the new §5 text — no collision. A2A sent.Origin: external Gemini review + @tobiu direction.