What is the 'Neural Link'?

The Neural Link is a bi-directional bridge that connects AI agents (like Gemini or Claude) directly to the Neo.mjs runtime. It allows agents to 'see' the application's Scene Graph, inspect component state, verify event listeners, and even mutate the running application in real-time. This turns the application into a 'glass box' for AI, enabling autonomous debugging and feature development.

Why is Neo.mjs called an 'Application Engine' instead of a framework?

Traditional frameworks are general-purpose libraries (like a Toyota) that help you organize code, but they compile away into ephemeral DOM nodes. Neo.mjs is a precision-engineered runtime (like an F1 car), similar to Unreal Engine for games. It maintains a persistent 'Scene Graph' of objects in a separate worker thread. These objects retain their identity, state, and relationships, allowing for advanced capabilities like multi-window orchestration, runtime permutation, and deep AI introspection that are impossible with 'melted plastic' DOM-based frameworks.

What is 'Context Engineering'?

Context Engineering is the practice of curating the environment and information flow for AI agents to maximize their autonomy. In Neo.mjs, this is implemented via three Model Context Protocol (MCP) servers: a Knowledge Base for semantic code understanding, a Memory Core for learning from past sessions, and a GitHub Workflow server for project management. This ecosystem allows agents to work as fully integrated members of the development team.

What is 'Object Permanence' in the context of Neo.mjs?

In Neo.mjs, UI components are persistent JavaScript objects living in the App Worker, not just transient rendering results. This 'Object Permanence' means a component (like a Dashboard) maintains its state (scroll position, user input, internal logic) even if it is detached from the DOM or moved to a different browser window. This is the 'Lego Technic' approach versus the 'Duplo' approach of traditional frameworks.

What is an 'Agent Operating System'?

Neo.mjs v11 introduced the concept of an Agent OS, where the platform itself provides the tools and interfaces for AI agents to operate. It combines a standalone, type-safe AI SDK for autonomous 'Code Execution' with the Neural Link for runtime control. This enables agents to monitor, debug, and heal the application autonomously, effectively acting as an operating system for synthetic intelligence.

How does Neo.mjs handle multi-window applications?

Neo.mjs uses shared web workers to run a single application instance across multiple browser windows. All windows share the same application state and data in real-time. Components can even move between windows while retaining their JavaScript instances. This enables desktop-class experiences like multi-window IDEs, browser-based email clients, and multi-screen control rooms.

What makes Neo.mjs different from React, Angular, or Vue?

Neo.mjs is fundamentally different in its architecture: (1) It uses True Multithreading via web workers to prevent UI jank, (2) It is an AI-Native Application Engine with built-in bridges for AI collaboration, and (3) It treats components as persistent objects in a Scene Graph, enabling native multi-window support and runtime permutation.

Frontmatter

id	11425
title	Separate ADR 0005 approval semantics from human merge gate
state	Closed
labels	documentationenhancementaiarchitecturemodel-experience
assignees	neo-gpt
createdAt	May 15, 2026, 3:07 PM
updatedAt	May 15, 2026, 3:35 PM
githubUrl	https://github.com/neomjs/neo/issues/11425
author	neo-gpt
commentsCount	0
parentIssue	null
subIssues	[]
subIssuesCompleted	0
subIssuesTotal	0
blockedBy	[]
blocking	[]
closedAt	May 15, 2026, 3:35 PM

Separate ADR 0005 approval semantics from human merge gate

Name: Neo.mjs Application Engine
Author: Neo.mjs

Closeddocumentationenhancementaiarchitecturemodel-experience

neo-gpt commented on May 15, 2026, 3:07 PM

Context

ADR 0005 currently conflates two different control surfaces: peer/maintainer PR approval and the human-only merge/content gate. The friction surfaced during 2026-05-15 PR review coordination when an agent withheld approval because operator content approval was not yet present. Operator clarification: peer maintainers may approve PRs when their criteria pass; @tobiu then performs an additional content/risk/bigger-picture pass before human merge.

Verified evidence:

learn/agentos/decisions/0005-adr-at-graduation-for-ideation-sandbox.md:48-53 names the section "merge-gate boundary" but says implementation PRs MUST NOT be approved/merged until the ADR is Accepted.
learn/agentos/decisions/0005-adr-at-graduation-for-ideation-sandbox.md:188-192 defines Accepted as requiring operator content-accuracy confirmation.
.agents/skills/post-review-pickup/references/post-review-pickup-workflow.md:27 already treats Approve as moving the PR to the human merge gate, not as executing or replacing the human gate.
.agents/skills/blocked-task-state/references/blocked-task-state-workflow.md:36 separately names completed merge eligibility as a PR with approvals waiting for the human merge gate.

The Problem

ADR 0005's current wording makes approved and merged look like one combined forbidden state. That creates a closed loop:

Agents interpret ADR 0005 as blocking their formal GitHub approval until operator content approval exists.
Operator content/risk review normally happens after the team has approved and the PR reaches the human merge gate.
The PR can get stuck with agents waiting for a human signal that is intended to happen after agent approval, not before it.

This weakens the flat maintainer model. @neo-gpt, @neo-opus-4-7, and @neo-gemini-3-1-pro are repo maintainers who can approve PRs on their own criteria. Human merge authority remains separate and absolute.

The Architectural Reality

ADR 0005 is still right that ADR-required implementation must not merge against unresolved authority. The bug is narrower: it over-constrains reviewer approval semantics.

Correct boundary:

Peer approval gate: agents approve when structural, test, evidence, and workflow criteria pass.
Human merge/content gate: @tobiu performs final content/risk/bigger-picture review and executes merge.
ADR-required sequencing: merge remains blocked until the relevant ADR state is acceptable at merge time. This may be represented by Status: Accepted, explicit operator content approval, or an approved ADR-update PR merged first, depending the concrete lane.

The human gate is not a prerequisite for peer approval; it is the next gate after peer approval.

The Fix

Amend ADR 0005 and any companion review-map wording that copied the conflation. The amendment should preserve ADR-required merge discipline while replacing approved/merged wording with separated approval and merge semantics.

Likely surfaces:

learn/agentos/decisions/0005-adr-at-graduation-for-ideation-sandbox.md
.agents/skills/pr-review/references/pr-review-guide.md if it says reviewers must block approval rather than block merge eligibility
.agents/skills/ideation-sandbox/references/ideation-sandbox-workflow.md if the same approval/merge wording landed there

Contract Ledger Matrix

Target Surface	Source of Authority	Proposed Behavior	Fallback / Edge Case	Docs	Evidence
ADR 0005 approval/merge boundary	ADR 0005, operator clarification 2026-05-15, #11425	Separate peer/team approval from human merge/content gate. Agents may approve when their own review criteria pass; human merge gate remains separate.	If ADR authority is not acceptable for merge, review may approve structurally while explicitly noting human merge/content/ADR-sequencing gate remains. Missing human approval alone is not a `CHANGES_REQUESTED` reason.	Yes: ADR 0005 amended.	Diff removes combined `approved/merged` semantics and adds separated approval/merge language.
PR review guidance for ADR-required work	ADR 0005 and `.agents/skills/pr-review/references/pr-review-guide.md`	Reviewer verifies ADR authority as a merge-readiness gate, not as a blanket peer-approval blocker.	Real review blockers still require `CHANGES_REQUESTED`; unresolved human merge/content gate is reported as merge-gate pending after approval.	Yes if matching wording exists.	Grep for `before approval`, `approved/merged`, and `ADR 0005` across review/ideation skill payloads.
Ideation graduation guidance for ADR-required work	ADR 0005 and `.agents/skills/ideation-sandbox/references/ideation-sandbox-workflow.md`	Implementation merge remains blocked until ADR authority is accepted; planning and review progression are not blocked unnecessarily.	If no copied conflation exists, leave untouched and cite grep evidence in PR body.	Only if matching wording exists.	Targeted grep + diff review.

Acceptance Criteria

ADR 0005 no longer uses approved/merged as a single combined forbidden state.
ADR 0005 explicitly distinguishes peer/team approval from the human-only merge/content gate.
ADR-required implementation PRs remain merge-blocked until the governing ADR authority is acceptable at merge time.
Review guidance allows APPROVE when peer criteria pass, with review body/A2A noting any remaining human merge/content gate.
No language implies agent approval can substitute for human merge authorization.
No language implies missing operator merge/content approval alone is a reason for agent CHANGES_REQUESTED when the agent's own review criteria pass.

Out of Scope

No change to AGENTS.md invariant forbidding gh pr merge.
No change to @tobiu's final content/risk/bigger-picture guard.
No weakening of ADR-required authority sequencing for actual merges.

ADR 0005: learn/agentos/decisions/0005-adr-at-graduation-for-ideation-sandbox.md
PR review/post-review pickup semantics: .agents/skills/post-review-pickup/references/post-review-pickup-workflow.md
Human merge gate precedent: #10274

Handoff Retrieval Hints

Retrieval Hint: query_raw_memories("ADR 0005 approved merged operator content approval closed loop") Retrieval Hint: query_raw_memories("11371 ADR 0005 Draft Accepted approve operator content accuracy")

tobiu closed this issue on May 15, 2026, 3:35 PM

tobiu referenced in commit e8d5f49 - "docs(agentos): separate ADR approval and merge gates (#11425) (#11426) on May 15, 2026, 3:35 PM