diff --git a/CHANGELOG.md b/CHANGELOG.md
index 5c013c1c2..ec96e35fa 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -81,6 +81,17 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   `launch_colab: true`. The flat names are the nested paths these become once
   mystmd supports structured template options, so that migration is mechanical
   ([#205](https://github.com/QuantEcon/quantecon-theme.mystmd/issues/205)) ([#224](https://github.com/QuantEcon/quantecon-theme.mystmd/pull/224)).
+- `PLAN.md` is closed. It opens with a note saying the plan is complete as of
+  the release that closes the book-theme parity gate, that the rest is the
+  planning snapshot it was written as and is no longer maintained, and where to
+  look instead — `docs/` for current behaviour, the gate issue for the
+  item-level record. The file itself is kept, because issues link to its lines.
+  The three pieces of guidance that outlived the plan moved into
+  `CONTRIBUTING.md` first: the architecture constraint (the theme never sees
+  the source repository, so anything derived from it must arrive as data), the
+  upstream-first rule and `UPSTREAM-CANDIDATES.yml`, and how the visual gate
+  and the PR previews work
+  ([#218](https://github.com/QuantEcon/quantecon-theme.mystmd/issues/218)) ([#230](https://github.com/QuantEcon/quantecon-theme.mystmd/pull/230)).
 
 ### Fixed
 - Images in a notebook cell's outputs are centred in the content column, as
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
index f5f740a15..19c6f88e0 100644
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@@ -60,6 +60,46 @@ public/          # Static assets (logos, Thebe bundles)
 patches/         # patch-package patches for upstream fixes
 ```
 
+## Architecture: what the theme can and cannot compute
+
+The two QuantEcon themes run in fundamentally different places, and one
+consequence of that shapes most feature work here.
+
+| | `quantecon-book-theme` (Sphinx) | this theme (MyST) |
+| --- | --- | --- |
+| Runs | at **build time**, inside the lecture repo | as a **runtime Remix server**, over pre-built content JSON |
+| Sees the git repo? | **yes** — it can shell out to `git` | **no** — only each page's `mdast` and `frontmatter` |
+| Customised through | Jinja templates and a Sphinx extension | React components and `myst.yml` |
+
+**So any feature derived from the source repository cannot be computed by the
+theme.** Git history, last-modified dates and computed launch paths all have to
+reach it as data, put there by one of:
+
+- a **MyST plugin or transform** that runs during `myst build` in the lecture
+  repo and writes the value into the page's frontmatter — `plugins/git-metadata.mjs`
+  is the one this repo ships;
+- **built-in mystmd support**, where it already exists (frontmatter `date`, say);
+- a **CI step** that pre-computes the value and feeds it in.
+
+The component's job is then only to render what it was given. When a feature
+looks impossible here, this is usually why, and the fix belongs upstream of the
+theme.
+
+## Upstream first, then here
+
+This theme tracks upstream
+[`jupyter-book/myst-theme`](https://github.com/jupyter-book/myst-theme)'s `book`
+theme. **Before building anything custom, check whether it already exists
+upstream** — or whether it could be contributed there rather than kept as
+QuantEcon-only code.
+
+Where it does not, build it here first: nothing waits on an upstream release.
+Record what could later be upstreamed in
+[`UPSTREAM-CANDIDATES.yml`](./UPSTREAM-CANDIDATES.yml), the feature-level
+registry, noting the local PR as provenance. This repo is not a fork of
+`jupyter-book/myst-theme`, so a candidate has to be ported rather than
+cherry-picked.
+
 ## Making Changes
 
 1. **Create a branch** from `main`:
@@ -158,9 +198,10 @@ accurate; a comment that narrates a project phase goes stale once the work is
 done.
 
 This applies to comments in code — including the comments in YAML, shell and
-fixture files. Prose documentation (`README.md`, `docs/`, `PLAN.md`, the test
-suite's own `README.md`) is written for a reader who wants the project's
-history, and keeps it.
+fixture files. Prose documentation (`README.md`, `docs/`, the test suite's own
+`README.md`) is written for a reader who wants the project's history, and keeps
+it — as does `PLAN.md`, which is a closed planning record rather than a live
+one.
 
 - **No project framing.** No phase numbers, `PLAN.md` items or milestones, and
   no issue number used as a label for the work that produced the code.
@@ -207,6 +248,29 @@ sites. Open an issue instead (see #172 for the code palette, #201 for the
 footer). When you compute a contrast ratio, composite any `opacity` on the
 element first: the declared colour is not what the reader sees.
 
+## CI: visual tests and PR previews
+
+Two things run on every pull request.
+
+**The visual gate.** A `visual` job pixel-diffs the fixture (desktop, mobile and
+sidebar-open) against the committed baselines in
+`tests/visual/__snapshots__`. Baselines are platform-suffixed: refresh the
+`-linux` ones by commenting `/update-snapshots` on the PR, and the `-darwin`
+ones locally with `npm run test:visual:update`. Playwright is the only gate.
+
+**The rendered preview.** `preview.yml` builds a real lecture site —
+`QuantEcon/lecture-python-programming` — with the PR's theme, through a static
+`myst build --html`. That is the export path a live `myst start` never
+exercises, so the preview catches static-build defects the Playwright suite
+cannot. It deploys to the `gh-pages` branch under `pr-preview/pr-<n>/`,
+sticky-comments the link, and tears the directory down when the PR closes. It
+needs no org secrets, only `GITHUB_TOKEN`.
+
+The content repo is still a legacy Jupyter Book, so the workflow's `myst init`
+step doubles as a migration-readiness check: if the Jupyter Book upgrade breaks
+on real lecture content, the preview build is where it shows. The preview is
+qualitative — it is not a gate.
+
 ## Commit Convention
 
 We use conventional commits:
diff --git a/PLAN.md b/PLAN.md
index 517314fe1..4077958d8 100644
--- a/PLAN.md
+++ b/PLAN.md
@@ -1,5 +1,24 @@
 # PLAN — QuantEcon MyST theme feature parity with `quantecon-book-theme`
 
+> **Closed.** This plan is complete as of the release that closes
+> [#147](https://github.com/QuantEcon/quantecon-theme.mystmd/issues/147), the
+> book-theme parity cutover gate. Everything below is the planning snapshot it
+> was written as, and is **no longer maintained** — the gap table in particular
+> records what was true when each phase was planned, not what the theme does
+> now.
+>
+> - For **what the theme does today**, read [`docs/`](docs/), starting at
+>   [`docs/index.md`](docs/index.md).
+> - For the **item-level record** of the parity work — every gate item, its
+>   decision and the PR that closed it — read
+>   [#147](https://github.com/QuantEcon/quantecon-theme.mystmd/issues/147) and
+>   the 2026-09-11 parity validation study it summarises.
+> - The guidance here that outlived the plan has moved into
+>   [`CONTRIBUTING.md`](CONTRIBUTING.md): the architecture constraint, the
+>   upstream-first rule, and how the PR previews work.
+>
+> The file is kept because issues link to its lines.
+
 This plan tracks bringing the MyST theme (this repo, `quantecon-theme.mystmd` —
 formerly `quantecon-theme-src`, and no longer bundled to the now-archived
 [`QuantEcon/quantecon-theme`](https://github.com/QuantEcon/quantecon-theme) —