Blog post: the repositories split, and the cockpit we build next

Context

The operator's direction today (2026-10-09, relayed by the lead): the 13.2 notes and posts must tell the repositories split and what we are working on, the Fleet Manager. In the operator's words, "a release without impressive release notes, screenshots, blog post(s) and maybe the apps/workstation video is pointless".

The 13.2 posts so far:

• #19492 (#9854), the developer post on Dock Layouts and multi-window. In #14790's Phase 1 it is the flagship's "how it's built" piece.
• This ticket, the second post: the organism's story. It covers why one repository became five, and the product the split made room for, Agent Institution. #14790's Phase 3 schedules "the institution angle" as the last story beat. The operator's direction moves it up to the release itself.

Readers arrive at the Engine (3,288 stars, 248 forks). Agent Institution's repository has 0 forks, and all 11 of its stars come from the team and its accounts (GitHub API, 2026-10-09 13:31Z). A post is the strongest bridge the release has between the two.

The Problem

The split is the release's first hero. Told from inside one repository, it reads as a subtraction: things that moved away. No public artifact tells a developer:

• why the Engine, the Brain and the cockpit now release on separate lines;
• what each of the five homes is for;
• that the cockpit, Agent Institution, is an application built on this release's Engine, with the same Dock Layouts, multi-window and SharedWorkers the Dock post explains.

The notes' new chapter (filed alongside) is a pointer. A post can tell the why with receipts.

The Architectural Reality

• Sources:
  • #19491's split chapter (Ada), which mines the why, how and what with dates;
  • the Engine README's map of the five repository homes (#17812);
  • NEO_INLINE_CODE_0, the organism map;
  • the Institution's README and NEO_INLINE_CODE_1.
• Institution facts (Institution dev@7ae86bd):
  • Built on 13.2's dock. The cockpit imports the Engine's dock interaction and persistence (apps/agentos/view/fleet/cockpit/Container.mjs), and pop-out, tear-out and return live in VesselContainer.mjs.
  • SharedWorkers on: useSharedWorkers: true in apps/agentos/neo-config.json.
  • Where it stands: version 0.1.0. The browser quickstart boots without a Brain and shows a cold "not answered yet" state until a Fleet transport answers.
• Bar: the blog authoring guide:
  • a thesis TL;DR, the reader's real problem, receipts and a CTA;
  • every external claim sourced, and the five over-claim flavors audited;
  • at least two model reviews, one of them cross-family, with the operator approving last;
  • the post registered in apps/portal/resources/data/blog.json.
• Product boundary: local-first framing only. #15519 puts "commercial-tier framing on any public surface" out of scope. Its sequencing truth keeps the public actions to watching and following until a download works.

The Fix

learn/blog/<slug>.md plus its blog.json leaf, in four beats:

1. The reader's problem. One repository carried an engine, the AI team that maintains it, and that team's cockpit. What that cost a consumer installing neo.mjs, and a contributor cloning it, uses #19491's measured facts.
2. How the split ran: the homes and removals in order, with dates, and an honest bound on what is unfinished.
3. The five homes: one Mermaid map, render-verified.
4. What we build next. Agent Institution: what it is, that it runs on the Engine the reader just installed (the bridge to #19492), where it stands, and how to follow it.

Acceptance Criteria

• AC-1 The post is in learn/blog/ with a blog.json leaf; the JSON parses and the link guard passes.
• AC-2 It opens with a thesis TL;DR and the reader's problem, and closes with a CTA to watch or star the Institution and to install the Engine.
• AC-3 Every factual claim links its primary. The split facts match #19491's merged chapter, and the Institution facts cite its dev at the writing SHA.
• AC-4 One Mermaid map of the repository homes renders in the portal, with proof in the PR body.
• AC-5 The five over-claim flavors are audited, with no hosted, paid or pricing framing and no promise D#19493 has not settled.
• AC-6 At least two model reviews, one of them cross-family. The operator approves last and decides when it publishes.

Out of Scope

• FM v1 decisions (D#19493).
• The FM outward door: storefront, site and download (#15519).
• Cross-posting after publication.
• The notes' chapter (its own ticket).

Avoided Traps

• The subtraction frame. The split is told as what it made possible, not as what left the Engine.
• A launch post for a cockpit a stranger can't run yet. The post tells what we build and where it stands; the run-it door belongs to #15519.

Related

#19491 (the split chapter) · #19492 / #9854 (the Dock post) · #14790 (the launch playbook) · #15519 (the FM outward door) · D#19493 (FM v1) · #14800 (the notes)

Live latest-open sweep: the latest 20 open issues, created-descending, at 2026-10-09T13:33Z, plus a title search for Agent Institution / what is next. No equivalent: #19491 is the split chapter only, #15519 / #15522 / #15520 are the FM outward door (storefront, naming), #14790 the playbook. A2A in-flight sweep: all-states mailbox, last 30, 12:37–13:31Z. The 12:55Z day board lists "the blog post(s)" as unowned; no claim on the notes frame or a post since. Memory Core sweep (problem nouns: blog post, repositories split, Fleet Manager, what we build next, visibility): no prior post decision; #13383 (the v13 blog-post epic) is closed; #9850, #9849 and #9853 are other subjects. Own-assignment sweep: #9854 is the Dock post, a different subject; no duplicate.

Retrieval Hint: "13.2 second blog post repositories split Agent Institution what we build next"

Origin Session ID: e76b2469-377c-4fec-85a7-4c47b10269b9