chore: clean up content phase organization

chore(config): add GitHub repository metadata to docs config
2026-11-02 13:12:44 +00:00 · 2026-09-25 14:00:22 -04:00 · 2026-09-25 13:36:50 -04:00 · 2026-09-25 13:30:06 -04:00 · 2026-09-25 13:27:06 -04:00 · 2026-09-25 13:17:35 -04:00
218 changed files with 7391 additions and 2371 deletions
@@ -114,7 +114,12 @@ One issue per curriculum topic; owns the problem sub-issues.
  ```

 `Scheduled` should come from the calendar of record — `apps/api/data/schedule.json`
-(day → topic issue) — not guessed.
+(`date → { topics: [issue, …] }`; a doubled day lists two) — not guessed.
+
+Gate problems (`gate: [...]` in the same file) must exist in the catalog before their
+Saturday, or the Worker refuses to create the review issue: create each as a
+`set:optional` problem sub-issue of its natural topic. Being named in the plan keeps it
+out of the drill pool, so it stays unseen until the gate.

 ## Reference tables

@@ -0,0 +1,150 @@
+---
+name: problem-approach
+description: Write the `## Approach` section — a Mermaid flowchart that traces the submitted solution — for every docs solution page under apps/docs/content/ that lacks one, using the repo's exact placement and node conventions. Use when asked to add/implement/fill in missing approaches, draw the approach diagram for a problem page, or when `/problem-approach` is run.
+---
+
+# Problem approach diagrams
+
+Every solution page (`apps/docs/content/(<category>)/<num>-<slug>.mdx`) has three
+hand-owned sections around the script-owned one:
+
+```
+frontmatter → <Badge> → ::::warning (maybe) → ### Example N / ### Constraints
+## Approach        ← this skill (Mermaid flowchart of the code below)
+## Solution        ← bun run sync owns this; never edit inside it
+## Explanation     ← bun run videos --write (NeetCode embed); not this skill
+```
+
+`sync.ts` regenerates only `## Solution`, so an `## Approach` placed **before** it
+survives every re-sync; anything placed inside `## Solution` is overwritten on the
+next `work/` push.
+
+## The command: fill every missing approach
+
+1. **Find the pages** (numbered pages with no `## Approach` heading):
+
+   ```sh
+   grep -L '^## Approach' apps/docs/content/*/*.mdx
+   ```
+
+2. **For each page**, read `## Solution` and write the diagram from *that code* — the
+   one on the page, not the textbook algorithm. The reader compares diagram and code
+   line by line; the diagram is a trace, not a lecture.
+
+3. **Insert** the section between the `### Constraints:` list and `## Solution`,
+   blank-line separated on both sides:
+
+   ````md
+   - `-100 <= Node.val <= 100`
+
+   ## Approach
+
+   ```mermaid
+   flowchart TD
+     …
+   ```
+
+   ## Solution
+   ````
+
+4. **Check** each diagram by hand against the syntax rules below (the docs build does
+   not parse Mermaid — a broken diagram fails only in the browser), then run the
+   docs build once at the end and review the pages in `bun run dev`:
+
+   ```sh
+   cd apps/docs && bunx blume build --isolated && rm -rf .blume-verify
+   ```
+
+Pages with two languages (`<CodeGroup>`) get **one** diagram, traced from the
+Python version. Pages with only JavaScript trace the JavaScript.
+
+## Diagram conventions
+
+Reference page: `apps/docs/content/(array)/15-3sum.mdx`. Every existing diagram
+follows this shape:
+
+```mermaid
+flowchart TD
+  S(["threeSum(nums)"]) --> O["nums.sort() — duplicates become adjacent"]
+  O --> F{"more i in 0..n-1?"}
+  F -- no --> E(["return result"])
+  F -- yes --> D{"nums[i] == nums[i-1]?"}
+  D -- yes --> F
+  D -- no --> P["left = i+1, right = n-1, target = -nums[i]"]
+  P --> W{"left < right?"}
+  W -- no --> F
+  W -- yes --> C["current = nums[left] + nums[right]"]
+  C --> Q{"current vs target"}
+  Q -- equal --> A["append triplet, skip equal neighbours, left += 1, right -= 1"]
+  A --> W
+  Q -- "current < target" --> L["left += 1 — need a bigger sum"]
+  L --> W
+  Q -- "current > target" --> R["right -= 1 — need a smaller sum"]
+  R --> W
+```
+
+- `flowchart TD`, two-space indent, one edge per line, edges listed in execution
+  order (entry first, exits as soon as they are reachable).
+- **Entry** is a stadium holding the signature: `S(["threeSum(nums)"])`. **Every
+  exit** is a stadium: `E(["return result"])`, `R(["return true"])`.
+- **Statements** are rectangles `["…"]`, in the code's own identifiers and operators
+  (`left += 1`, `freq[n] = (freq[n] or 0) + 1`). Several trivial assignments may share
+  one node, comma-separated.
+- **Decisions** are diamonds `{"…?"}` with `-- yes -->` / `-- no -->` edges. A
+  three-way branch names the outcomes: `-- equal -->`, `-- "current < target" -->`
+  (quote an edge label whenever it holds `<`, `>`, or spaces around operators).
+- **Loops** are a decision: `for` → `{"more n in nums?"}`, `while` → the condition
+  verbatim `{"left < right?"}`. `-- no -->` goes to whatever follows the loop; the
+  body's last node edges back to the decision. `continue` is an edge back to the loop
+  decision (`D -- yes --> F` above); `break` is an edge to the node after the loop.
+- **Recursion** is a statement, not an expanded subgraph: `A["left = invertTree(root.left)"]`.
+  The base case is the first decision after entry.
+- **Classes** (design problems) use one `subgraph name["method(args)"]` per public
+  method, plus a `state["…"]` subgraph describing the fields, with `state --> method`
+  edges — see `(hash-table)/146-lru-cache.mdx`. Private helpers are named inside the
+  method nodes (`"_remove_node then _add_node — move it beside the tail"`), not drawn.
+- **Annotate the why with an em dash** inside the node, sparingly (one clause, a few
+  per diagram): `"nums.sort() — duplicates become adjacent"`, `"map = {0: 1} — the
+  empty prefix"`. The dash explains the invariant that makes the step correct — the
+  one thing the code does not say.
+- **Node ids**: one or two capitals, unique per diagram (`S I L Q C E R W A B`). Class
+  diagrams prefix per method (`G1`, `P1`).
+
+### Syntax rules (a diagram that breaks the page is worse than none)
+
+- Wrap **every** label in double quotes — brackets, parentheses, `<`, `>` and `|` in an
+  unquoted label break the parser.
+- Never put `"` inside a label; use `'` for string literals (`"s == ''?"`).
+- Never use `|` anywhere — not in node labels, not as `-->|text|` edge syntax; edge
+  labels are always `-- text -->`. Write `or` for a logical or.
+- Every id an edge references must be declared with a shape the first time it appears.
+- Balanced: each `subgraph` has an `end`; each fence opens ```` ```mermaid ```` and closes ```` ``` ````.
+
+### Code that does not run
+
+Trace it anyway — the diagram documents what was submitted — and put a `::::warning`
+callout **before** the fence naming exactly why it fails (`(array)/121-best-time-to-buy-and-sell-stock.mdx`):
+
+````md
+## Approach
+
+::::warning
+The diagram traces the code below literally. That code does not run: `curr` and `ans` are never initialised, so the first `curr > left` test raises `NameError`.
+::::
+
+```mermaid
+…
+```
+````
+
+Do not fix the code on the page: `## Solution` mirrors `work/1`, and the fix belongs
+there (then `bun run sync`).
+
+## Do not
+
+- Do not write prose approaches, complexity tables or alternative algorithms — the
+  section is one flowchart (plus the warning callout when the code is broken).
+- Do not touch `## Solution` or anything after it; `## Explanation` is filled by
+  `bun run videos --write`.
+- Do not add an approach to `reference/` pages or `index.mdx` — only numbered
+  solution pages (`title: 'NN. …'`) carry one.
@@ -0,0 +1,36 @@
+#!/bin/sh
+# Pre-commit: the docs pages under apps/docs/content (and the `/<num>` short
+# links in apps/docs/redirects.ts) are generated from work/1 by `bun run sync`,
+# and deploy.yml publishes whatever main holds. Regenerate them now and stage
+# what sync changed, so a solution, its page and its redirect land in the same
+# commit and a stale `## Solution` never reaches main.
+#
+# Sync reads the working tree, not the index, so the pages it writes only
+# match this commit when work/1 has nothing unstaged: an untracked or
+# half-staged solution would ship a page whose source is not in the commit.
+# In that case refuse and say what to stage.
+#
+# Only paths that became dirty during sync are staged. Anything already dirty
+# beforehand (hand-written `## Approach` prose in progress, a new index.mdx)
+# stays where the author left it.
+#
+# Bypass for a deliberate partial commit: git commit --no-verify
+set -eu
+
+cd "$(git rev-parse --show-toplevel)"
+
+unstaged=$(git status --porcelain -- work/1 | grep -v '^[MADR] ' || true)
+if [ -n "$unstaged" ]; then
+  printf 'pre-commit: work/1 has unstaged changes; stage or stash them so the generated pages match this commit:\n%s\n' "$unstaged" >&2
+  exit 1
+fi
+
+before=$(mktemp)
+trap 'rm -f "$before"' EXIT
+git status --porcelain -- apps/docs/content apps/docs/redirects.ts | sort > "$before"
+report=$(bun run --silent sync)
+generated=$(git status --porcelain -- apps/docs/content apps/docs/redirects.ts | sort | comm -13 "$before" - | cut -c4-)
+if [ -n "$generated" ]; then
+  printf '%s\n' "$report" | grep -v 'skipped (in sync)' >&2
+  printf '%s\n' "$generated" | xargs git add --
+fi
@@ -7,7 +7,7 @@ on:
  workflow_dispatch:
    inputs:
      dry_run:
-        description: Report matches without closing anything
+        description: Report matches and would-be releases without closing or releasing anything
        type: boolean
        default: false

@@ -61,3 +61,39 @@ jobs:
            exit 0
          fi
          bun apps/cli/sync-project-fields.ts
+
+  # One timever release per finished topic (every core sub-issue closed),
+  # tagged at the instant the last core problem closed. It reads the same rule
+  # close-topics uses (apps/cli/topics.ts) straight off the sub-issues, so the
+  # push that finishes a topic releases it in this run — no waiting for Close
+  # Topics, which chains after this workflow. Reconciling: a run releases
+  # every finished topic still missing one, so a manual dispatch backfills and
+  # a problem closed by hand in the UI is picked up by the next run.
+  release:
+    needs: close
+    # Like close-topics: a partially failed close still closed some issues,
+    # and the script only reads state, so run unless the run was cancelled.
+    if: ${{ !cancelled() }}
+    runs-on: ubuntu-latest
+    permissions:
+      contents: write # push tags, create releases
+      issues: read
+    steps:
+      # Full history: each tag targets the last main commit before its topic
+      # finished, and backfilled tags reach far behind HEAD.
+      - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - uses: oven-sh/setup-bun@v2
+
+      # The tagger of the annotated, backdated tags.
+      - run: |
+          git config user.name "github-actions[bot]"
+          git config user.email "41898282+github-actions[bot]@users.noreply.github.com"
+
+      # No install step: the script only uses Bun builtins + fetch.
+      - run: bun apps/cli/release-topics.ts
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          DRY_RUN: ${{ inputs.dry_run }}
@@ -30,7 +30,13 @@ jobs:
      - uses: oven-sh/setup-bun@v2 # skip if you use npm/pnpm

      - run: bun install --frozen-lockfile
+      # blume.config.ts pulls the /changelog pages from this repo's GitHub
+      # releases (githubReleases source) at build time. The token authenticates
+      # that fetch — required if the repo is private, and it lifts the shared
+      # runner's unauthenticated API rate limit. `contents: read` covers it.
      - run: bun run build
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

      - uses: actions/upload-pages-artifact@v3
        with:
@@ -14,6 +14,7 @@ flowchart LR
  work -->|close-solved.yml| PI[problem issues]
  work -->|close-solved.yml POST /admin/solved| D1[(D1 = SRS state)]
  PI -->|close-topics.yml| TI[topic issues]
+  PI -->|close-solved.yml release job| Rel[timever release per finished topic]
  GH[GitHub issues = catalog] -->|nightly reconcile| D1
  D1 -->|8 AM ET cron| Mail[digest email + one-tap links]
  Mail -->|GET /log| D1
@@ -33,9 +34,11 @@ flowchart LR
 - **The picker owns the local review loop; the buckets are its state.** `work/` holds three buckets (`1` = first solve, `3`/`7` = blind re-solves, named after the ladder rungs) and `bun run pick` has three sections — Tab cycles them. Pools are reconciled from the filesystem plus one `git log` (`apps/cli/work.ts`): a first solve dates from the ET date of the commit that added its `work/1` file — the same push that closes the problem issue and starts the +3/+7 windows — and a window stops being owed the moment its bucket holds a file. Uncapped, unlike the review issue's `WINDOW_CAP`: that issue is one day's assignment, the picker is everything still owed. An empty scaffold under `work/1` owes nothing (it is unsolved), which is why `isImplemented()` lives in `apps/cli/source.ts` and is shared with `close-solved.ts`. leetcode-cli reads its output directory from `~/.leetcode/workspaces/<active>/config.json` and nowhere else — no flag, no env var — so `pick.ts` MERGES `workDir` into that file around the child (never replays saved bytes: the CLI owns the file too) and leaves it pointed at `work/1`, then treats the scaffold's existence as the acceptance test, since leetcode-cli exits 0 even when an expired session made it write nothing. The picker itself writes no state: pushing the re-solve is what moves the ladder, via `close-solved` → `/admin/solved` with that file's bucket.
 - **The README charts are live, never committed.** They are Worker endpoints reading D1 per request (`Cache-Control: max-age=300`, honored by GitHub Camo), so a solved push moves them within ~5 minutes with no commit and no workflow of their own.
 - **D1 owns SRS state; GitHub issues own the catalog; `apps/api/data/schedule.json` owns the calendar.** The catalog reconcile never invents rows and never overwrites SRS columns (`stage`, `next_review`). `Target Date` is the *only* Project field the Worker mirrors (`apps/api/src/mirror.ts`): failures are warnings, never lost D1 writes; topic rows' `Target Date` is never written. SRS stage and first-attempt result live **only** in D1 — do not re-add them as Project fields. The Project's `Set`/`Difficulty` single-selects are a projection of the issue labels, reconciled by `bun run sync-project-fields` — a best-effort final step of `close-solved.yml`, never the Worker.
+- **The plan names days; the samplers fill the gaps.** `apps/api/data/schedule.json` maps a date to a `ScheduledDay` (`apps/api/src/srs.ts`): `topics` (one or more topic issues — a doubled day lists two, and the digest renders each with its core list then the optional "if time" list), `problems` (a set day: fixed numbers, rendered like a topic), `gate` (Saturday's blind set, verbatim — `createReviewIssue` refuses to create the issue if any number is not a catalog row, so gate problems must exist as `set:optional` sub-issues of their topic first), or `sweep` (every problem whose last attempt failed, taps to re-log, no drills). Every number a `gate` or `problems` entry names is in `RESERVED`, and the drill/gate samplers never draw it — that is what keeps a planned gate problem unseen until its Saturday. Dates absent from the plan still get the ladder's reviews and drills; a Saturday without a `gate` entry falls back to sampling this week's optional pools.
 - **Determinism = idempotency.** Drill/gate sampling uses a seeded PRNG (`rng()` in `apps/api/src/srs.ts`, FNV-1a → mulberry32, seed = date / ISO week); the digest is keyed by ET date in `email_log`; one-tap links are unique on (problem, date, kind).
 - **DST-proof crons:** each event has two UTC crons; code fires only on the computed ET hour (`etHour`), unit-checked in `apps/api/src/srs.test.ts`.
 - **Chained workflows:** `GITHUB_TOKEN`-driven issue closes fire no `issues` events, so `close-topics.yml` chains off `workflow_run` of Close Solved instead. The Worker's webhook uses its own `GH_PAT`, so its events flow normally.
+- **One "finished topic" rule, two consumers.** `apps/cli/topics.ts` (`coreSet()`: every `set:core` sub-issue closed, zero core = never) is read by `close-topics.ts` (closes the topic issue) and `release-topics.ts` (the `release` job of `close-solved.yml`). The release reads the sub-issues, not the topic issue, so the push that finishes a topic releases it in the same run instead of waiting for Close Topics to chain after it. Version = [timever](https://gist.github.com/twolfson/de1b004dd22536b8e668) (`YYYYMMDD.HHMMSS.NNNNNNNNN`, UTC) of the instant the last core problem closed; the tag is annotated and backdated to that instant, on the last first-parent main commit before it; the body is a changelog snapshot at that instant (a one-sentence summary first — Blume's changelog makes it the page's meta description — then closed on the finish day / earlier / open at release, with solution links at the tag). A release is found by its `<!-- release-topics: #N -->` body marker, never by tag name, so re-closing a core problem never mints a second one; an existing release keeps its tag and only has its body rewritten when the render changes. The docs `/changelog` is these releases (`githubReleases` in `apps/docs/blume.config.ts`, fetched at build with `GITHUB_TOKEN`), dated by `published_at` — which the API cannot backdate, so the Aug 24 – Sep 24 backfill reads Sep 25 there, and its true finish time lives in the version and the summary.

 ## Key Directories

@@ -43,9 +46,10 @@ flowchart LR
 |---|---|
 | `work/<1\|3\|7>/<Difficulty>/<Category>/<num>.<slug>.{js,py}` | Solutions, e.g. `work/1/Easy/Array/1.two-sum.py`. The top directory is the spaced-repetition bucket: `1` = first solve, `3`/`7` = the blind re-solves of that problem 3 and 7 days later. Machine-parsed header comment (title / `Difficulty:` / URL / `─` rule / statement) — preserve its exact shape |
 | `apps/cli/` | Flat Bun TS workspace: automation entries (shebang + top-level await) and libraries (no shebang, side-effect-free on import). Root `bun run` scripts delegate here |
-| `apps/docs/` | Blume site (`blume.config.ts`, `content/`, `islands/`, `public/`). `content/(<category>)/<num>-<slug>.mdx` generated by sync |
-| `apps/api/` | Cloudflare Worker workspace: `src/` modules, `data/schedule.json` (day → topic issue, human-edited, bundled at deploy), `migrations/` (append-only; `0002` is the +3/+7 ladder rebuild) |
-| `.github/workflows/` | deploy (docs → Pages, Worker → Cloudflare, on every main push), close-solved (issues + `/admin/solved` + Project `Set`/`Difficulty`, needs `SRS_ADMIN_KEY` and `PROJECT_PAT`), close-topics, sync-d1 (issue edits → `/admin/reconcile`), spaced-repetition (daily 06:00 ET cron → today's review issue; `GITHUB_TOKEN` only, no secrets) |
+| `apps/docs/` | Blume site (`blume.config.ts`, `content/`, `components/` + `components.ts`, `islands/`, `public/`). `content/(0N-<phase>)/<topic>/<num>-<slug>.mdx` plus one `meta.ts` per phase and per topic folder, all generated by sync: the phase folder is a `(group)` (sidebar section, no route segment), the topic folder is the README campaign topic slug (`tree-dfs/`, `heap-priority-queue/`) and a real route segment (`/tree-dfs/invert-binary-tree`), the `meta.ts` files carry the exact labels, phase/topic numbers and icons so the sidebar reads in curriculum order with each topic a collapsible group, and the page badge is the `work/` data-structure category. A hand-written `(0N-<phase>)/<topic>/index.mdx` (`hidden: true`) is that topic's landing page: Blume ≥ 1.7.3 links the sidebar group row to it (`sliding-window/index.mdx` → `/sliding-window`) and the hidden flag drops the duplicate row. Two kinds of custom MDX component, both import-free on every page: `islands/<Name>.tsx` is hydrated React (auto-discovered; `ProgressDashboard`, `DependencySpine` — the README topic map as data, zoom/pan, nodes link to whatever route Blume's `/api/docs/navigation.json` reports for the topic folder, inert until a topic has pages), `components/<Name>.astro` registered in `components.ts` is static markup plus a custom element for behaviour, the way Blume's own `Frame` is built (`Carousel`/`CarouselImage`: `src` takes a URL, `/file.png`, or the page-relative `../../public/file.png`) |
+| `apps/api/` | Cloudflare Worker workspace: `src/` modules, `data/schedule.json` (the plan: `date → { topics | problems | gate | sweep }`, human-edited, bundled at deploy), `migrations/` (append-only; `0002` is the +3/+7 ladder rebuild) |
+| `.github/workflows/` | deploy (docs → Pages with `GITHUB_TOKEN` for the releases-backed changelog, Worker → Cloudflare, on every main push), close-solved (issues + `/admin/solved` + Project `Set`/`Difficulty`, needs `SRS_ADMIN_KEY` and `PROJECT_PAT`; then the `release` job: timever tag + release per finished topic, `contents: write`), close-topics, sync-d1 (issue edits → `/admin/reconcile`), spaced-repetition (daily 06:00 ET cron → today's review issue; `GITHUB_TOKEN` only, no secrets) |
+| `.githooks/` | Tracked git hooks; `bun install` points `core.hooksPath` here (root `prepare` script, no husky). `pre-commit` runs `bun run sync` and stages only what sync changed (pages, `apps/docs/redirects.ts`), so a solution, its page and its `/<num>` short link land in one commit; it refuses to run while `work/1` has anything unstaged (sync reads the tree, so the page would not match the commit), and leaves pre-existing dirt under `apps/docs/content` (in-progress `## Approach` prose) unstaged. `git commit --no-verify` for a deliberate partial commit |

 ## Development Commands

@@ -65,10 +69,15 @@ bun run test            # run ONE solution against LeetCode's judge (not a test
                        # most recently tested first, stamped in the db on exit 0)
 bun run submit          # submit ONE solution to LeetCode (newest scaffold first, nothing
                        # hidden; last_submitted stamped in the db on exit 0)
-bun run sync            # work/1 → apps/docs/content/ pages (re-solves are never published)
+bun run sync            # work/1 → apps/docs/content/ pages + apps/docs/redirects.ts (/<num> short
+                        # links; re-solves are never published)
+bun run videos          # docs pages with no <YouTube> embed → NeetCode video per page
+                        # (channel search scraped, no API key; --write appends the
+                        # `## Explanation` embed; "no NeetCode match" pages stay untouched)
 bun run dev|build       # Blume docs site (runs in apps/docs/)
 bun run close-solved -- --dry-run    # issue reconcilers (also DRY_RUN=1)
 bun run close-topics -- --dry-run
+bun run release-topics -- --dry-run # prints every would-be topic release body
 # close-solved also pushes the solved set to the Worker; point it at a local
 # one and nothing production is touched:
 SRS_API=http://localhost:8787 SRS_ADMIN_KEY=$LINK_KEY bun run close-solved -- --dry-run
@@ -101,11 +110,11 @@ curl -X POST -H "Authorization: Bearer $LINK_KEY" \
 - `apps/api/src/index.ts` — router + cron dispatch; `apps/api/wrangler.jsonc` — bindings (`DB`, `EMAIL`), crons, vars; secrets `GH_PAT`/`WEBHOOK_SECRET`/`LINK_KEY` via `wrangler secret put`.
 - `apps/api/src/digest.ts` / `apps/api/src/email.tsx` — the daily digest, split data/presentation. `digest.ts` reads D1 into a `DigestData`; `email.tsx` owns every colour and every sentence, and renders both the HTML and (via its own `plainDigest`, not React Email's `plainText` mode, which flattens the tables) the text alternative. The retrieval rules hold by construction: `DigestRow` has no title and no issue field, so a review or drill line *cannot* leak the topic or a solution link. Subject is `(Day N/56) LeetCode Daily Digest`.
 - `apps/api/src/png.ts` — hand-rolled PNG encoder (RGB8, one IDAT, zlib via `CompressionStream("deflate")`, CRC32, 5×7 bitmap font). It exists because every major email client refuses remote SVG, so `/chart/heatmap.png` rasters the heatmap for the digest while `/chart/heatmap.svg` keeps serving the README byte-for-byte. Both come from one `heatmapCells()` so the two pictures cannot drift.
- `apps/cli/sync.ts` — work→docs contract: on existing pages only the `## Solution` section itself is script-owned, bounded by `solutionEnd()` at the next unfenced top-level heading; human prose before it and any sections after it (`## Approach`, `## Explanation`) are never touched. Never hand-write solution pages or edit inside `## Solution`.
+- `apps/cli/sync.ts` — work→docs contract: on existing pages only the `<Badge>` line and the `## Solution` section are script-owned, the latter bounded by `solutionEnd()` at the next unfenced top-level heading; human prose before it and any sections after it are never touched. Never hand-write solution pages or edit inside `## Solution`. Two classifications, two homes: the sidebar group is the README topic (`readmeTopics()`, first listing wins for a problem under two topics; a warm-up outside the plan groups under its category at `order: 0`, ahead of topic 1) and the badge is the `work/1/<Difficulty>/<Category>/` category. Topics nest under their README phase (`## Phase N — Name` headings, shortened via `PHASE_ALIASES`): `content/(0N-<phase>)/<topic>/`, both `meta.ts` script-owned and rewritten every run (phase: `Phase N: Name`, `order: N`; topic: label, curriculum order, `TOPIC_ICONS` icon, `display: "group"`). The phase folder is a Blume `(group)` — no route segment, so `/tree-dfs` is unchanged and a landing page there is one level deep, never at `/` — and the topic folder is the real segment; never put a topic `index.mdx` directly inside a `(group)`. A page is `rename`d when its topic or phase changes; a folder every page has left is deleted (climbing into an emptied phase folder) unless it still holds a hand-written `index.mdx` (sync never writes, moves, or deletes one — so page-relative asset paths in a moved landing page are yours to fix). Slugs come from `slugify()`, which glues a leading number to the next token (`1d-dynamic-programming`) because Blume strips `<digits>-` prefixes from routes. The two hand-owned sections around the solution have their own owners: `## Approach` (a Mermaid flowchart tracing the page's code, placed *before* `## Solution` so a re-sync keeps it) is written per the `problem-approach` skill (`/problem-approach` fills every page missing one); `## Explanation` (the NeetCode `<YouTube>` embed, after `## Solution`) is appended by `bun run videos --write` (`apps/cli/videos.ts`).
 - `apps/cli/work.ts` — the `work/` layout: bucket paths, `isSolution()`/`solvedInBucket()` (the stub test, shared with `close-solved.ts`), `firstSolved()` (one `git log`, ET dates, keyed by LC number so past layout moves don't matter), and `reviewQueues()` (what each window still owes). Everything that needs to know where a solution file lives, or when it was solved, goes through here.
 - `apps/api/src/stats.ts` / `apps/cli/stats.ts` / `apps/cli/tui.ts` / `apps/cli/api.ts` — the read side. `buildStats()` is the ONE aggregation: phases, ladder, difficulty, per-topic temperature, gates, the 14-day due queue, and `heat` (every campaign day, zeros and future days included — the docs page windows the tail off it rather than asking a second question). Two readers consume it, the docs island and `bun run stats`, so its shape is both their contract: change all three together. `tui.ts` is hand-rolled ANSI (bars, sparklines, heat cells, alternate screen) on charts.ts's palette, and every chart differs by GLYPH as well as by colour so a piped or `NO_COLOR` dashboard still reads. The TUI recomputes nothing: the only local fact it shows is the `work/` solution count, which is inventory, not schedule.
 - `README.md` live charts are Worker endpoints (`/chart/*.svg`, `/badge/gate.svg`, 5-min Camo cache); the docs `/progress` page fetches `/api/stats` client-side (`apps/docs/islands/ProgressDashboard.tsx`).
- `apps/docs/blume.config.ts` — site base `/leetcode`; `README.md` campaign table doubles as topic-map input to `sync.ts` (`readmeTopics()`, `TOPIC_ALIASES`).
+- `apps/docs/blume.config.ts` — site base `/leetcode`; imports `apps/docs/redirects.ts`, sync-generated (`/<num>` → `/<topic>/<slug>`, one row per page, exact root-relative paths — Blume adds the base and, for the static Pages build, emits a meta-refresh page per rule plus `blume-redirects.json`); `README.md` campaign table doubles as topic-map input to `sync.ts` (`readmeTopics()`, `TOPIC_ALIASES`, `PHASE_ALIASES`).
 - Old `.github/srs/` JSON state, its `srs-*` Actions, and the one-shot `import-srs.ts` importer are RETIRED and deleted — do not resurrect; git history is the record.

 ## Runtime/Tooling Preferences
@@ -4,8 +4,8 @@

 <p align="center">
  <a href="https://x.com/basedprad"><img alt="follow" src="https://shieldcn.dev/x/follow/basedprad.svg" /></a>
-  <a href="https://lc.prad.nu"><img alt="badge" src="https://shieldcn.dev/badge/Read-Documentation.svg?logo=lu%3ABook" /></a>
-  <a href="https://lc.prad.nu/progress"><img alt="gate pass rate" src="https://srs-api.prdlk.workers.dev/badge/gate.svg" /></a>
+  <a href="https://leetcode.com/progress"><img alt="Custom badge" src="https://shieldcn.dev/badge/view-@prdlk.svg?logo=leetcode" /></a>
+  <a href="https://lc.prd.lk"><img alt="badge" src="https://shieldcn.dev/badge/Read-Documentation.svg?logo=lu%3ABook" /></a>
 </p>

 <p align="center">
@@ -337,4 +337,3 @@
    <tr><td align="center"><code>B</code></td><td align="center">252</td><td><a href="https://leetcode.com/problems/meeting-rooms/">Meeting Rooms I</a></td></tr>
  </tbody>
 </table>
-
@@ -14,8 +14,12 @@ Live at `https://srs-api.prdlk.workers.dev`.
  labels, milestones. `catalog.ts` reconciles them into D1 nightly (and via
  `/admin/reconcile`); it recomputes from scratch, never invents rows, and
  never overwrites SRS-owned columns (`stage`, `next_review`).
- **The repo owns the schedule** — `data/schedule.json`, human-edited,
-  bundled at deploy. Git history is its audit log.
+- **The repo owns the plan** — `data/schedule.json`, human-edited,
+  bundled at deploy. Git history is its audit log. Each date is a
+  `ScheduledDay`: `topics` (one or more topic issues), `problems` (a set day,
+  fixed numbers), `gate` (Saturday's blind set, verbatim), or `sweep` (redo
+  every miss, nothing new). Numbers a `gate` or `problems` entry names are
+  `RESERVED`: the drill and gate samplers never draw them.
 - The GitHub Project mirror writes `Target Date` and nothing else: it is
  best-effort, so failures log warnings and never block a D1 write, and topic
  rows are never written. SRS stage and first-attempt result stay in D1 only —
@@ -47,7 +51,10 @@ raw `Date` math.
 Blind drills draw only from topics **already learned**: scheduled in an
 earlier week (the current week's optional pool is reserved for Saturday's
 gate) and with at least one core problem on the ladder — a skipped learning
-day never feeds drills just because its calendar week lapsed.
+day never feeds drills just because its calendar week lapsed. A Saturday
+whose plan fixes the `gate` set uses that list as-is; every number must be a
+catalog row (a `set:optional` sub-issue of its topic), or the review issue is
+not created — a `/done` on an uncatalogued number could never be logged.

 ## Routes

@@ -1,26 +1,28 @@
 {
-  "2026-08-17": 3,
-  "2026-08-18": 4,
-  "2026-08-24": 5,
-  "2026-08-25": 6,
-  "2026-08-26": 7,
-  "2026-08-27": 8,
-  "2026-08-28": 9,
-  "2026-08-31": 10,
-  "2026-09-01": 11,
-  "2026-09-02": 12,
-  "2026-09-03": 13,
-  "2026-09-04": 14,
-  "2026-09-07": 15,
-  "2026-09-08": 16,
-  "2026-09-09": 17,
-  "2026-09-10": 18,
-  "2026-09-11": 19,
-  "2026-09-14": 20,
-  "2026-09-15": 21,
-  "2026-09-16": 22,
-  "2026-09-17": 23,
-  "2026-09-18": 24,
-  "2026-09-21": 25,
-  "2026-09-22": 26
+  "2026-08-17": { "topics": [3] },
+  "2026-08-18": { "topics": [4] },
+  "2026-08-24": { "topics": [5] },
+  "2026-08-25": { "topics": [6] },
+  "2026-08-26": { "topics": [7] },
+  "2026-08-27": { "topics": [8] },
+  "2026-08-28": { "topics": [9] },
+  "2026-08-31": { "topics": [10] },
+  "2026-09-01": { "topics": [11] },
+  "2026-09-02": { "topics": [12] },
+  "2026-09-03": { "topics": [13] },
+  "2026-09-04": { "topics": [14] },
+  "2026-09-07": { "topics": [15] },
+  "2026-09-08": { "topics": [16] },
+  "2026-09-21": { "topics": [17, 18] },
+  "2026-09-22": { "topics": [19, 20] },
+  "2026-09-23": { "topics": [21] },
+  "2026-09-24": { "topics": [22] },
+  "2026-09-25": { "topics": [23] },
+  "2026-09-26": { "gate": [648, 1091, 802, 547, 90] },
+  "2026-09-28": { "topics": [24] },
+  "2026-09-29": { "topics": [25] },
+  "2026-09-30": { "topics": [26] },
+  "2026-10-01": { "problems": [399, 261, 909, 77, 52] },
+  "2026-10-02": { "sweep": true },
+  "2026-10-03": { "gate": [347, 1254, 1971, 131, 494, 986] }
 }
@@ -5,14 +5,16 @@
 * the HTML and the plain-text alternative from that single tree (React Email
 * `render`), so the two can never drift.
 *
- * Retrieval rules: review and drill rows carry number + difficulty only —
- * never the topic, never a solution link. `DigestRow` has no field for
- * either, so the rule holds by construction. The learning day's core list is
- * the only labeled section. At most REVIEW_CAP (3) reviews per day; drills
- * fill toward DAILY_CAP (6) only when the due queue is light. Overflow is
- * levelled, not stacked: sendDigest gives every review past the cap a
- * concrete future date, ≤ REVIEW_CAP per day, oldest first. Sunday is a
- * two-line rest note.
+ * Retrieval rules: review, drill and sweep rows carry number + difficulty
+ * only — never the topic, never a solution link. `DigestRow` has no field for
+ * either, so the rule holds by construction. The plan's own sections — a
+ * learning day's topics (core, then the optional "if time" list) and a set
+ * day's problems — are the only labeled, linked ones. At most REVIEW_CAP (3)
+ * reviews per day; drills fill toward DAILY_CAP (6) only when the due queue
+ * is light, and never on a sweep day, which assigns every open miss and
+ * nothing new. Overflow is levelled, not stacked: sendDigest gives every
+ * review past the cap a concrete future date, ≤ REVIEW_CAP per day, oldest
+ * first. Sunday is a two-line rest note.
 *
 * Idempotency: email_log keys sends by ET date — a same-day re-send is a
 * no-op unless forced; the body itself is deterministic (drill picks are
@@ -22,20 +24,23 @@
 import {
  CAMPAIGN_DAYS,
  type ProblemRow,
+  RESERVED,
  REVIEW_CAP,
  SCHEDULE,
+  type ScheduledDay,
  addDays,
  campaignDay,
  campaignWeek,
  dueReviews,
  isoWeek,
  levelReviews,
+  lookupProblems,
  pickDrills,
  prettyDate,
  streak,
  weekdayOf,
 } from "./srs.ts";
-import { type DigestData, type DigestRow, renderDigest } from "./email.tsx";
+import { type DigestData, type DigestRow, type TopicRow, renderDigest } from "./email.tsx";
 import { signLink } from "./links.ts";

 /** Total daily ceiling: drills only fill what the whole due queue leaves. */
@@ -63,23 +68,50 @@ async function tapRow(env: Env, p: ProblemRow, date: string, staged: boolean): P
  };
 }

+/** A plan-named problem: linked to its issue, ticked once it has entered the ladder. */
+function topicRow(env: Env, p: ProblemRow): TopicRow {
+  return {
+    lc: p.lc_number,
+    difficulty: p.difficulty,
+    url: `https://github.com/${env.REPO}/issues/${p.issue}`,
+    solved: p.stage !== "new",
+  };
+}
+
+/** Problems whose most recent attempt failed — the miss log a sweep day works through. */
+async function openMisses(db: D1Database): Promise<ProblemRow[]> {
+  const { results } = await db
+    .prepare(
+      `SELECT p.* FROM problems p
+       JOIN attempts a ON a.lc_number = p.lc_number
+       WHERE a.id = (SELECT MAX(id) FROM attempts WHERE lc_number = p.lc_number)
+         AND a.result = 'fail' AND p.stage != 'retired'
+       ORDER BY a.date, p.lc_number`,
+    )
+    .all<ProblemRow>();
+  return results;
+}
+
 /** Everything the email needs, read straight out of D1 and the schedule. */
 export async function collectDigest(env: Env, date: string): Promise<DigestData> {
  const db = env.DB;
  const rest = weekdayOf(date) === 0;
+  const plan: ScheduledDay = rest ? {} : (SCHEDULE[date] ?? {});

  // Reviews surface at most REVIEW_CAP; drills fill toward DAILY_CAP only
  // when the WHOLE due queue is light — a heavy backlog must never trade
-  // review slots for brand-new drill problems.
+  // review slots for brand-new drill problems — and never on a sweep day.
  const due = rest ? [] : await dueReviews(db, date);
  const capped = due.slice(0, REVIEW_CAP);
-  const drills = rest ? [] : await pickDrills(db, date, DAILY_CAP - Math.min(due.length, DAILY_CAP));
+  const drills =
+    rest || plan.sweep ? [] : await pickDrills(db, date, DAILY_CAP - Math.min(due.length, DAILY_CAP));

  const data: DigestData = {
    day: prettyDate(date),
    progress: `Day ${campaignDay(date)} of ${CAMPAIGN_DAYS} · Week ${campaignWeek(date)}`,
    streak: await streak(db, date),
    rest,
+    topics: [],
    reviews: await Promise.all(capped.map((p) => tapRow(env, p, date, true))),
    carried: due.length - capped.length,
    drills: await Promise.all(drills.map((p) => tapRow(env, p, date, false))),
@@ -88,28 +120,42 @@ export async function collectDigest(env: Env, date: string): Promise<DigestData>
    progressUrl: `${env.DOCS_URL}/progress`,
  };

-  // New topic — the only section allowed to name problems and link them.
-  const topicIssue = rest ? undefined : SCHEDULE[date];
-  if (topicIssue !== undefined) {
+  // New topics — the plan's sections, the only ones allowed to name problems
+  // and link them. Core first; the optional set is "if time", minus anything
+  // the plan has reserved for a gate or a set day.
+  for (const topicIssue of plan.topics ?? []) {
    const topic = await db
      .prepare("SELECT name FROM topics WHERE issue = ?")
      .bind(topicIssue)
      .first<{ name: string }>();
-    const { results: core } = await db
+    const { results } = await db
      .prepare(
-        "SELECT * FROM problems WHERE topic_issue = ? AND set_label = 'core' ORDER BY lc_number",
+        "SELECT * FROM problems WHERE topic_issue = ? AND set_label IN ('core', 'optional') ORDER BY lc_number",
      )
      .bind(topicIssue)
      .all<ProblemRow>();
-    data.topic = {
+    data.topics.push({
      name: topic?.name ?? `#${topicIssue}`,
-      core: core.map((p) => ({
-        lc: p.lc_number,
-        difficulty: p.difficulty,
-        url: `https://github.com/${env.REPO}/issues/${p.issue}`,
-        solved: p.stage !== "new",
-      })),
-    };
+      core: results.filter((p) => p.set_label === "core").map((p) => topicRow(env, p)),
+      ifTime: results
+        .filter((p) => p.set_label === "optional" && !RESERVED.has(p.lc_number))
+        .map((p) => topicRow(env, p)),
+    });
+  }
+
+  // A set day assigns problems by number rather than by topic.
+  if (plan.problems) {
+    const { rows, missing } = await lookupProblems(db, plan.problems);
+    if (missing.length) console.warn(`digest ${date}: planned problems not in the catalog: ${missing.join(", ")}`);
+    data.assigned = rows.map((p) => topicRow(env, p));
+  }
+
+  // A sweep day works the miss log: every problem still standing on a
+  // failure, minus the ones today's review list already carries.
+  if (plan.sweep) {
+    const reviewed = new Set(capped.map((p) => p.lc_number));
+    const misses = (await openMisses(db)).filter((p) => !reviewed.has(p.lc_number));
+    data.sweep = await Promise.all(misses.map((p) => tapRow(env, p, date, true)));
  }

  // Saturday: the review issue already exists (created at midnight ET).
@@ -83,7 +83,7 @@ export interface DigestRow {
  failUrl: string;
 }

-/** A learning-day core problem — the only rows allowed a link. */
+/** A plan-named problem — the only rows allowed a link. */
 export interface TopicRow {
  lc: number;
  difficulty: string;
@@ -91,6 +91,13 @@ export interface TopicRow {
  solved: boolean;
 }

+/** One learning-day topic: the core list, then the optional "if time" list. */
+export interface TopicSection {
+  name: string;
+  core: TopicRow[];
+  ifTime: TopicRow[];
+}
+
 export interface DigestData {
  /** "Tuesday, August 25" */
  day: string;
@@ -98,11 +105,16 @@ export interface DigestData {
  progress: string;
  streak: number;
  rest: boolean;
-  topic?: { name: string; core: TopicRow[] };
+  /** Topics taught today, in plan order — empty on every other day. */
+  topics: TopicSection[];
+  /** A set day's problems, assigned by number rather than by topic. */
+  assigned?: TopicRow[];
  reviews: DigestRow[];
  /** Reviews past the daily cap; levelled onto the coming days. */
  carried: number;
  drills: DigestRow[];
+  /** A sweep day's miss log: present (even empty) only on a sweep day. */
+  sweep?: DigestRow[];
  gate?: { week: number; url: string };
  yesterday: { total: number; failed: number[] };
  gateRate?: number;
@@ -278,10 +290,15 @@ function Block({
 // ── email ────────────────────────────────────────────────────────

 function DigestEmail({ data }: { data: DigestData }) {
-  const load = data.reviews.length + data.drills.length;
-  const preview = data.rest
-    ? "Rest day — nothing to do but rest."
-    : `${load} to work through today${data.topic ? `, starting with ${data.topic.name}` : ""}.`;
+  const load = data.reviews.length + data.drills.length + (data.sweep?.length ?? 0);
+  const lead = data.topics[0]
+    ? `, starting with ${data.topics[0].name}`
+    : data.assigned
+      ? ", starting with today's set"
+      : data.sweep
+        ? " — sweep day, nothing new"
+        : "";
+  const preview = data.rest ? "Rest day — nothing to do but rest." : `${load} to work through today${lead}.`;

  const passed = data.yesterday.total - data.yesterday.failed.length;
  const missed = data.yesterday.failed.map((lc) => `LC ${lc}`).join(", ");
@@ -310,15 +327,47 @@ function DigestEmail({ data }: { data: DigestData }) {

            {data.rest ? null : (
              <>
-                {data.topic ? (
-                  <Section style={{ marginTop: "28px" }}>
-                    <Text style={label}>Something new today</Text>
+                {data.topics.map((topic, i) => (
+                  <Section key={topic.name} style={{ marginTop: i === 0 ? "28px" : "20px" }}>
+                    {i === 0 ? (
+                      <Text style={label}>
+                        {data.topics.length === 1 ? "Something new today" : "Two new things today"}
+                      </Text>
+                    ) : null}
                    <Text style={{ color: INK, fontSize: "16px", fontWeight: 600, margin: "0 0 8px" }}>
-                      {data.topic.name}
+                      {topic.name}
                    </Text>
                    <Text style={hint}>Work through these from the top — a tick means it is already solved.</Text>
-                    <TopicTable rows={data.topic.core} />
+                    <TopicTable rows={topic.core} />
+                    {topic.ifTime.length > 0 ? (
+                      <>
+                        <Text style={{ ...hint, margin: "12px 0 8px" }}>
+                          If time — only once the core is done inside the clock.
+                        </Text>
+                        <TopicTable rows={topic.ifTime} />
+                      </>
+                    ) : null}
                  </Section>
+                ))}
+
+                {data.assigned ? (
+                  <Section style={{ marginTop: "28px" }}>
+                    <Text style={label}>Today's set</Text>
+                    <Text style={hint}>
+                      Picked by number, not by topic — all of them, in order, inside the clock.
+                    </Text>
+                    <TopicTable rows={data.assigned} />
+                  </Section>
+                ) : null}
+
+                {data.sweep ? (
+                  <Block
+                    title="Sweep — redo every red"
+                    note="Every problem still standing on a miss. Nothing new today; clear these first."
+                    rows={data.sweep}
+                    staged
+                    empty="The miss log is clear — nothing is standing on a fail."
+                  />
                ) : null}

                <Block
@@ -409,16 +458,27 @@ function plainDigest(data: DigestData): string {
          : ` · ${data.streak} day${data.streak === 1 ? "" : "s"} in a row, keep it going.`),
  ];

+  const linked = (p: TopicRow) =>
+    `  LC ${p.lc} — ${DIFFICULTY[p.difficulty]?.label ?? p.difficulty}${p.solved ? " (done)" : ""} — ${p.url}`;
+
  if (!data.rest) {
-    if (data.topic) {
+    for (const topic of data.topics) {
      out.push(
        "",
-        `SOMETHING NEW TODAY — ${data.topic.name}`,
+        `SOMETHING NEW TODAY — ${topic.name}`,
        "Work through these from the top; the ones marked done you have already solved.",
-        ...data.topic.core.map(
-          (p) =>
-            `  LC ${p.lc} — ${DIFFICULTY[p.difficulty]?.label ?? p.difficulty}${p.solved ? " (done)" : ""} — ${p.url}`,
-        ),
+        ...topic.core.map(linked),
+      );
+      if (topic.ifTime.length > 0) {
+        out.push("  If time — only once the core is done inside the clock:", ...topic.ifTime.map(linked));
+      }
+    }
+    if (data.assigned) {
+      out.push(
+        "",
+        "TODAY'S SET",
+        "Picked by number, not by topic — all of them, in order, inside the clock.",
+        ...data.assigned.map(linked),
      );
    }

@@ -436,6 +496,14 @@ function plainDigest(data: DigestData): string {
        "No cold starts today.",
      ],
    ];
+    if (data.sweep) {
+      lists.unshift([
+        "SWEEP — REDO EVERY RED",
+        "Every problem still standing on a miss. Nothing new today; clear these first.",
+        data.sweep,
+        "The miss log is clear — nothing is standing on a fail.",
+      ]);
+    }
    for (const [title, note, rows, empty] of lists) {
      out.push("", title);
      if (rows.length === 0) {
@@ -2,21 +2,27 @@
 * Saturday review issue: created at midnight ET so the 8 AM digest can link
 * to it; scored when the issue closes (webhook).
 *
- * The gate half is a topic-blind quiz — one unsolved optional problem per
- * topic scheduled this week (which is why daily drills only draw from
- * earlier weeks). The recap half is generated data, deliberately NOT a task
- * list: the week's solved problems are already scheduled by the +3/+7 ladder,
- * and re-assigning them on Saturday would be massed practice.
+ * The gate half is a topic-blind quiz. When the plan fixes the set
+ * (`gate` in data/schedule.json) that list IS the gate, verbatim — every
+ * number must be a catalog problem, or the issue is not created, because a
+ * `/done` on an uncatalogued number can never be logged. Otherwise the set
+ * is sampled: one unsolved optional problem per topic scheduled this week
+ * (which is why daily drills only draw from earlier weeks), never one the
+ * plan has reserved. The recap half is generated data, deliberately NOT a
+ * task list: the week's solved problems are already scheduled by the +3/+7
+ * ladder, and re-assigning them on Saturday would be massed practice.
 *
 * Label is `review` (the retired Actions system owned `gate`). Creation is
- * idempotent by week; deferred Hards never appear.
+ * idempotent by week; deferred Hards never appear in a sampled set.
 */
 import {
  type ProblemRow,
+  RESERVED,
  type Result,
  SCHEDULE,
  campaignWeek,
  isoWeek,
+  lookupProblems,
  rng,
  sample,
  streak,
@@ -32,7 +38,7 @@ function capitalized(difficulty: string): string {
  return difficulty[0]!.toUpperCase() + difficulty.slice(1);
 }

-/** Unsolved problems of a topic in a set — never attempted, never drilled. */
+/** Unsolved problems of a topic in a set — never attempted, drilled, or reserved. */
 async function freshPool(db: D1Database, topic: number, set: string): Promise<ProblemRow[]> {
  const { results } = await db
    .prepare(
@@ -44,7 +50,7 @@ async function freshPool(db: D1Database, topic: number, set: string): Promise<Pr
    )
    .bind(topic, set)
    .all<ProblemRow>();
-  return results;
+  return results.filter((p) => !RESERVED.has(p.lc_number));
 }

 export interface CreateReport {
@@ -67,40 +73,52 @@ export async function createReviewIssue(env: Env, gh: GitHub, date: string): Pro
    return { created: false, issue: existing.issue, reason: `${title} exists (#${existing.issue})` };
  }

-  // Topics scheduled this week; review-only weeks (Sep 23 onward) sample
-  // across every topic that still has unsolved optional problems.
-  let topics = Object.entries(SCHEDULE)
-    .filter(([d]) => campaignWeek(d) === week)
-    .sort(([a], [b]) => (a < b ? -1 : 1))
-    .map(([, t]) => t);
-  if (topics.length === 0) {
-    const { results } = await db
-      .prepare(
-        `SELECT DISTINCT topic_issue AS t FROM problems
-         WHERE set_label = 'optional' AND stage = 'new'
-           AND lc_number NOT IN (SELECT lc_number FROM drill_pool_used)
-         ORDER BY topic_issue`,
-      )
-      .all<{ t: number }>();
-    topics = sample(results.map((r) => r.t), MAX_GATE_PROBLEMS, rng(`gate-topics-${iso}`)).sort(
-      (a, b) => a - b,
-    );
-  }
-  if (topics.length === 0) return { created: false, reason: "no topics with unsolved pools" };
-
-  const random = rng(`gate-${iso}`);
+  // The plan's fixed set wins; otherwise sample this week's topics, and in a
+  // week without topics, across every topic that still has an unsolved pool.
  const picks: { p: ProblemRow; fallback: boolean }[] = [];
-  for (const topic of topics) {
-    const optional = await freshPool(db, topic, "optional");
-    if (optional.length > 0) {
-      picks.push({ p: sample(optional, 1, random)[0]!, fallback: false });
-      continue;
+  const planned = SCHEDULE[date]?.gate;
+  if (planned) {
+    const { rows, missing } = await lookupProblems(db, planned);
+    if (missing.length) {
+      return {
+        created: false,
+        reason: `planned gate names problems outside the catalog: LC ${missing.join(", LC ")}`,
+      };
    }
-    const core = await freshPool(db, topic, "core");
-    if (core.length > 0) picks.push({ p: sample(core, 1, random)[0]!, fallback: true });
-    // Both pools exhausted: the topic is fully solved; nothing to quiz.
+    picks.push(...rows.map((p) => ({ p, fallback: false })));
+  } else {
+    let topics = Object.entries(SCHEDULE)
+      .filter(([d]) => campaignWeek(d) === week)
+      .sort(([a], [b]) => (a < b ? -1 : 1))
+      .flatMap(([, day]) => day.topics ?? []);
+    if (topics.length === 0) {
+      const { results } = await db
+        .prepare(
+          `SELECT DISTINCT topic_issue AS t FROM problems
+           WHERE set_label = 'optional' AND stage = 'new'
+             AND lc_number NOT IN (SELECT lc_number FROM drill_pool_used)
+           ORDER BY topic_issue`,
+        )
+        .all<{ t: number }>();
+      topics = sample(results.map((r) => r.t), MAX_GATE_PROBLEMS, rng(`gate-topics-${iso}`)).sort(
+        (a, b) => a - b,
+      );
+    }
+    if (topics.length === 0) return { created: false, reason: "no topics with unsolved pools" };
+
+    const random = rng(`gate-${iso}`);
+    for (const topic of topics) {
+      const optional = await freshPool(db, topic, "optional");
+      if (optional.length > 0) {
+        picks.push({ p: sample(optional, 1, random)[0]!, fallback: false });
+        continue;
+      }
+      const core = await freshPool(db, topic, "core");
+      if (core.length > 0) picks.push({ p: sample(core, 1, random)[0]!, fallback: true });
+      // Both pools exhausted: the topic is fully solved; nothing to quiz.
+    }
+    if (picks.length === 0) return { created: false, reason: `every pool for week ${week} is exhausted` };
  }
-  if (picks.length === 0) return { created: false, reason: `every pool for week ${week} is exhausted` };

  // Recap: everything attempted this week, plus streak — data, not tasks.
  const monday = weekMonday(date);
@@ -154,7 +172,7 @@ export async function createReviewIssue(env: Env, gh: GitHub, date: string): Pro

  const milestone = await db
    .prepare("SELECT milestone FROM topics WHERE issue = ? AND milestone IS NOT NULL")
-    .bind(topics[0])
+    .bind(picks[0]!.p.topic_issue)
    .first<{ milestone: number }>();

  const created = (await gh.rest(`/repos/${gh.repo}/issues`, {
@@ -8,6 +8,8 @@ import { describe, expect, test } from "bun:test";

 import {
  INTERVAL,
+  RESERVED,
+  SCHEDULE,
  STAGES,
  WINDOWS,
  addDays,
@@ -21,6 +23,7 @@ import {
  sample,
  TEMPERATURES,
  temperatureOf,
+  topicDate,
  weekdayOf,
  workingDay,
 } from "./srs.ts";
@@ -125,6 +128,38 @@ describe("the rest day", () => {
  });
 });

+describe("the plan", () => {
+  test("never books the rest day, and gates only on Saturday", () => {
+    for (const [date, day] of Object.entries(SCHEDULE)) {
+      expect(isRestDay(date)).toBe(false);
+      if (day.gate) expect(weekdayOf(date)).toBe(6);
+      if (day.topics) expect(weekdayOf(date)).toBeLessThan(6);
+    }
+  });
+
+  test("a topic is taught once; a doubled day answers for both", () => {
+    const seen = new Set<number>();
+    for (const day of Object.values(SCHEDULE)) {
+      for (const topic of day.topics ?? []) {
+        expect(seen.has(topic)).toBe(false);
+        seen.add(topic);
+      }
+    }
+    expect(topicDate(17)).toBe("2026-09-21");
+    expect(topicDate(18)).toBe("2026-09-21");
+    expect(topicDate(999)).toBeUndefined();
+  });
+
+  test("every gate and set problem is reserved from the samplers", () => {
+    for (const day of Object.values(SCHEDULE)) {
+      for (const lc of [...(day.gate ?? []), ...(day.problems ?? [])]) {
+        expect(RESERVED.has(lc)).toBe(true);
+      }
+    }
+    expect(RESERVED.size).toBeGreaterThan(0);
+  });
+});
+
 describe("the ladder", () => {
  test("two rungs, named after the windows", () => {
    expect(STAGES).toEqual(["new", "+3", "+7", "retired"]);
@@ -126,12 +126,30 @@ export function prettyDate(date: string): string {

 // ── the schedule (bundled; humans edit apps/api/data/schedule.json) ───

-export const SCHEDULE: Record<string, number> = scheduleJson;
+/**
+ * One calendar day of the plan. Most days teach topics (one, or two when the
+ * plan doubles up); the rest are named by what they assign instead: a fixed
+ * gate set for Saturday's review issue, a fixed problem set (a "B-set" day),
+ * or a sweep — redo every miss, nothing new. Absent dates are unplanned:
+ * the ladder still runs, the digest just has nothing new to say.
+ */
+export interface ScheduledDay {
+  /** Topic issues taught that day, in the order the digest presents them. */
+  topics?: number[];
+  /** Problems assigned by number instead of by topic. */
+  problems?: number[];
+  /** Saturday: the blind set, fixed by the plan rather than sampled. */
+  gate?: number[];
+  /** Miss-log sweep: every problem whose last attempt failed, and no new ones. */
+  sweep?: boolean;
+}
+
+export const SCHEDULE: Record<string, ScheduledDay> = scheduleJson;

 /** Date the schedule teaches a topic on, if it teaches it at all. */
 export function topicDate(topic: number): string | undefined {
-  for (const [date, t] of Object.entries(SCHEDULE)) {
-    if (t === topic) return date;
+  for (const [date, day] of Object.entries(SCHEDULE)) {
+    if (day.topics?.includes(topic)) return date;
  }
  return undefined;
 }
@@ -142,6 +160,16 @@ export function topicWeek(topic: number): number | undefined {
  return date === undefined ? undefined : campaignWeek(date);
 }

+/**
+ * Every problem the plan names outright — a gate's blind set or a set day's
+ * list. The plan owns these: the samplers (daily drills, a sampled gate)
+ * never draw them, so a gate problem stays unseen until its Saturday and a
+ * set day's problems wait for their day.
+ */
+export const RESERVED: ReadonlySet<number> = new Set(
+  Object.values(SCHEDULE).flatMap((day) => [...(day.gate ?? []), ...(day.problems ?? [])]),
+);
+
 // ── deterministic sampling ───────────────────────────────────────

 /** FNV-1a → mulberry32: seeded PRNG so re-runs pick identical problems. */
@@ -248,6 +276,31 @@ export async function getProblem(db: D1Database, lc: number): Promise<ProblemRow
  return db.prepare("SELECT * FROM problems WHERE lc_number = ?").bind(lc).first<ProblemRow>();
 }

+/**
+ * Catalog rows for a list of numbers the plan names, in the plan's order,
+ * plus the numbers the catalog does not know — a planned problem without an
+ * issue can never be logged, so callers decide whether that is fatal.
+ */
+export async function lookupProblems(
+  db: D1Database,
+  lcs: number[],
+): Promise<{ rows: ProblemRow[]; missing: number[] }> {
+  if (lcs.length === 0) return { rows: [], missing: [] };
+  const { results } = await db
+    .prepare(`SELECT * FROM problems WHERE lc_number IN (${lcs.map(() => "?").join(",")})`)
+    .bind(...lcs)
+    .all<ProblemRow>();
+  const byNumber = new Map(results.map((p) => [p.lc_number, p]));
+  const rows: ProblemRow[] = [];
+  const missing: number[] = [];
+  for (const lc of lcs) {
+    const row = byNumber.get(lc);
+    if (row) rows.push(row);
+    else missing.push(lc);
+  }
+  return { rows, missing };
+}
+
 /**
 * What "due" means, in one place: not retired, scheduled on/before ?1, and
 * past its deferral if it has one. Every caller that asks "what is open on
@@ -358,8 +411,9 @@ export async function levelReviews(db: D1Database, date: string): Promise<number
 * scheduled in an earlier week (the current week's optional pool is reserved
 * for Saturday's gate) AND showing learning evidence: at least one of the
 * topic's core problems has entered the ladder. A skipped learning day never
- * feeds drills just because its calendar week lapsed. Never repeated,
- * seeded by date. Boosted topics contribute up to 2 extra.
+ * feeds drills just because its calendar week lapsed. Problems the plan names
+ * outright (RESERVED) are never drawn. Never repeated, seeded by date.
+ * Boosted topics contribute up to 2 extra.
 */
 export async function pickDrills(
  db: D1Database,
@@ -382,6 +436,7 @@ export async function pickDrills(
    )
    .all<ProblemRow & { boost: number }>();
  const pool = results.filter((p) => {
+    if (RESERVED.has(p.lc_number)) return false;
    const w = topicWeek(p.topic_issue);
    return w !== undefined && w < week;
  });
@@ -32,8 +32,8 @@
    "REPO": "prdlk/leetcode",
    "FROM_EMAIL": "srs@prdlk.com",
    "TO_EMAIL": "prnk28@gmail.com",
-    "DOCS_URL": "https://prdlk.github.io/leetcode",
-    "DOCS_ORIGIN": "https://prdlk.github.io,https://lc.prad.nu",
+    "DOCS_URL": "https://lc.prd.lk",
+    "DOCS_ORIGIN": "https://lc.prd.lk,https://prdlk.github.io",
    "PUBLIC_URL": "https://srs-api.prdlk.workers.dev"
  }

@@ -7,6 +7,7 @@
 * `set:optional` and `set:deferred` are extra credit. A topic is finished when
 * every one of its core sub-issues is closed; optional/deferred state is
 * ignored, matching the day rule in each topic body ("Core problems first").
+ * The rule lives in topics.ts, shared with release-topics.ts.
 *
 * Like close-solved.ts this reconciles state instead of reacting to an event
 * payload: re-runs are no-ops, backfilling needs no special casing, and only
@@ -19,6 +20,7 @@
 */
 import { github } from "./github.ts";
 import { markdownTable, printTable, writeStepSummary } from "./report.ts";
+import { coreSet, readProblems } from "./topics.ts";

 const DRY =
  process.argv.includes("--dry-run") ||
@@ -53,52 +55,13 @@ for await (const raw of gh.list(`/repos/${gh.repo}/issues?labels=topic&state=ope
 }
 topics.sort((a, b) => a.number - b.number);

-// ── core sub-issue state per topic ───────────────────────────────
-
-interface Core {
-  /** Core sub-issues, closed and open alike, lowest number first. */
-  all: number[];
-  /** The core sub-issues still open — non-empty means the topic stays open. */
-  pending: number[];
-}
-
-/**
- * Read a topic's core sub-issue state. The sub_issues payload carries the full
- * issue objects (state + labels), so no per-problem follow-up request is needed.
- */
-async function readCore(topic: number): Promise<Core> {
-  const all: number[] = [];
-  const pending: number[] = [];
-  for await (const raw of gh.list(`/repos/${gh.repo}/issues/${topic}/sub_issues`)) {
-    if (!raw || typeof raw !== "object") continue;
-    if (!("number" in raw) || typeof raw.number !== "number") continue;
-    if (!("state" in raw) || typeof raw.state !== "string") continue;
-    if (!("labels" in raw) || !Array.isArray(raw.labels)) continue;
-
-    const core = raw.labels.some(
-      (label) =>
-        label &&
-        typeof label === "object" &&
-        "name" in label &&
-        label.name === "set:core",
-    );
-    if (!core) continue;
-
-    all.push(raw.number);
-    if (raw.state === "open") pending.push(raw.number);
-  }
-  all.sort((a, b) => a - b);
-  pending.sort((a, b) => a - b);
-  return { all, pending };
-}
-
 // ── reconcile ────────────────────────────────────────────────────

 const report: string[][] = [];
 let closed = 0;

 for (const topic of topics) {
-  const core = await readCore(topic.number);
+  const core = coreSet(await readProblems(gh, topic.number));
  const progress = `${core.all.length - core.pending.length}/${core.all.length}`;

  // A topic with no core sub-issues has nothing to complete: never close it,
@@ -8,12 +8,14 @@
    "test": "bun ./test.ts",
    "submit": "bun ./submit.ts",
    "sync": "bun ./sync.ts",
+    "videos": "bun ./videos.ts",
    "close-solved": "bun ./close-solved.ts",
-    "close-topics": "bun ./close-topics.ts"
+    "close-topics": "bun ./close-topics.ts",
+    "release-topics": "bun ./release-topics.ts"
  },
  "devDependencies": {
    "@clack/prompts": "^1.7.0",
-    "@night-slayer18/leetcode-cli": "^3.5.0",
+    "@night-slayer18/leetcode-cli": "^3.5.2",
    "@types/bun": "^1.3.14"
  }
 }
@@ -0,0 +1,333 @@
+#!/usr/bin/env bun
+/**
+ * Publish one GitHub release per finished topic, versioned by timever
+ * (`YYYYMMDD.HHMMSS.NNNNNNNNN`, UTC — https://gist.github.com/twolfson/de1b004dd22536b8e668).
+ *
+ * "Finished" is topics.ts's rule — every core sub-issue closed — the same one
+ * close-topics.ts closes the topic issue on, and the version is the instant
+ * the LAST core problem closed. Keying off the sub-issues rather than the
+ * topic issue lets this run as the tail of Close Solved: the push that closes
+ * a topic's final core problem releases it in the same run, without waiting
+ * for close-topics to chain after it. The topic issue itself closes seconds
+ * later, on the same day.
+ *
+ * Each release is a changelog snapshot at that instant: the topic's problem
+ * sub-issues grouped as closed on the finish day (ET), closed earlier, and
+ * open at release, each with its set, difficulty, and the solution files as
+ * they stood in the tagged commit — the last main commit before the finish.
+ *
+ * The tag is annotated and backdated (tagger date = finish), so
+ * `git tag --sort=taggerdate` reads as the campaign's history even for tags
+ * minted by a backfill. The release's own published date cannot be backdated
+ * (the API has no field for it), and the docs changelog (githubReleases in
+ * apps/docs/blume.config.ts) dates entries by it — so a backfilled entry
+ * carries its real finish time in the version and the body's first sentence.
+ *
+ * Reconciles instead of reacting: every finished topic without a release gets
+ * one, so a first run backfills the whole campaign; an existing release keeps
+ * its tag and commit and only has its body rewritten when render() would say
+ * something else. Re-runs with nothing new are no-ops.
+ * A release is found by the marker comment in its body, never by tag name, so
+ * reopening and re-closing a core problem can never mint a second release.
+ * A tag that exists without its release (a run that died between the two) is
+ * reused, not re-created.
+ *
+ *   bun apps/cli/release-topics.ts            # tag, push, release, fix bodies
+ *   bun apps/cli/release-topics.ts --dry-run  # print the would-be releases
+ *
+ * Auth: GH_TOKEN / GITHUB_TOKEN, else `gh auth token` (needs contents: write).
+ * Git:  a clone with full history, an `origin` it may push tags to, and a
+ *       committer identity (the tagger).
+ */
+import { etDate } from "../api/src/srs.ts";
+import { github } from "./github.ts";
+import { markdownTable, printTable, writeStepSummary } from "./report.ts";
+import { coreSet, labelValue, readProblems, type Problem } from "./topics.ts";
+import { ROOT, parseWorkPath } from "./work.ts";
+
+const DRY =
+  process.argv.includes("--dry-run") ||
+  process.env.DRY_RUN === "1" ||
+  process.env.DRY_RUN === "true";
+
+/** The docs site (apps/docs/blume.config.ts `deployment.site`); sync publishes `/<num>` short links. */
+const DOCS_SITE = "https://lc.prd.lk";
+const TIMEVER = "https://gist.github.com/twolfson/de1b004dd22536b8e668";
+
+const gh = await github();
+
+function git(...args: string[]): string {
+  const run = Bun.spawnSync(["git", ...args], { cwd: ROOT });
+  if (run.exitCode !== 0) throw new Error(`git ${args.join(" ")}: ${run.stderr.toString().trim()}`);
+  return run.stdout.toString().trim();
+}
+
+// ── topics ───────────────────────────────────────────────────────
+
+interface Topic {
+  number: number;
+  title: string;
+  milestone: string | null;
+  body: string;
+}
+
+/** Narrow one /issues element; the `topic` label filter cannot exclude PRs. */
+function readTopic(value: unknown): Topic | undefined {
+  if (!value || typeof value !== "object") return;
+  if ("pull_request" in value) return;
+  if (!("number" in value) || typeof value.number !== "number") return;
+  if (!("title" in value) || typeof value.title !== "string") return;
+  const milestone =
+    "milestone" in value &&
+    value.milestone &&
+    typeof value.milestone === "object" &&
+    "title" in value.milestone &&
+    typeof value.milestone.title === "string"
+      ? value.milestone.title
+      : null;
+  const body = "body" in value && typeof value.body === "string" ? value.body : "";
+  return { number: value.number, title: value.title, milestone, body };
+}
+
+interface Finished {
+  topic: Topic;
+  problems: Problem[];
+  core: number[];
+  finishedAt: string;
+}
+
+// Closed and open topics alike: the rule reads the sub-issues, so a topic
+// finished moments ago (close-topics not yet run) releases now.
+const finished: Finished[] = [];
+for await (const raw of gh.list(`/repos/${gh.repo}/issues?labels=topic&state=all`)) {
+  const topic = readTopic(raw);
+  if (!topic) continue;
+  const problems = await readProblems(gh, topic.number);
+  const core = coreSet(problems);
+  if (core.finishedAt) finished.push({ topic, problems, core: core.all, finishedAt: core.finishedAt });
+}
+// Oldest first, so a backfill creates releases in campaign order.
+finished.sort((a, b) => a.finishedAt.localeCompare(b.finishedAt) || a.topic.number - b.topic.number);
+
+// ── existing releases ────────────────────────────────────────────
+
+/** Topic -> its release. The body is kept in sync with render() on every run. */
+const released = new Map<number, { id: number; tag: string; body: string }>();
+for await (const raw of gh.list(`/repos/${gh.repo}/releases`)) {
+  if (!raw || typeof raw !== "object") continue;
+  if (!("id" in raw) || typeof raw.id !== "number") continue;
+  if (!("tag_name" in raw) || typeof raw.tag_name !== "string") continue;
+  if (!("body" in raw) || typeof raw.body !== "string") continue;
+  const m = raw.body.match(/<!-- release-topics: #(\d+) -->/);
+  if (m) released.set(Number(m[1]), { id: raw.id, tag: raw.tag_name, body: raw.body });
+}
+
+// ── timever ──────────────────────────────────────────────────────
+
+/**
+ * Timever of an instant. The API reports whole seconds, so SUBTIME is the
+ * tiebreak: two topics finishing in the same second (one push closing both
+ * last core problems) get consecutive nanoseconds, in (instant, topic) order.
+ */
+const tags = new Map<number, string>(); // topic -> tag
+{
+  let prev = "";
+  let ns = 0;
+  for (const f of finished) {
+    const iso = new Date(f.finishedAt).toISOString(); // 2026-09-24T18:36:38.000Z
+    const base = `${iso.slice(0, 10).replaceAll("-", "")}.${iso.slice(11, 19).replaceAll(":", "")}`;
+    ns = base === prev ? ns + 1 : Number(iso.slice(20, 23)) * 1_000_000;
+    prev = base;
+    tags.set(f.topic.number, `${base}.${String(ns).padStart(9, "0")}`);
+  }
+}
+
+// ── release body ─────────────────────────────────────────────────
+
+const ET_WHEN = new Intl.DateTimeFormat("en-US", {
+  timeZone: "America/New_York",
+  weekday: "short",
+  month: "short",
+  day: "numeric",
+  year: "numeric",
+  hour: "numeric",
+  minute: "2-digit",
+});
+const ET_DAY = new Intl.DateTimeFormat("en-US", {
+  timeZone: "America/New_York",
+  month: "short",
+  day: "numeric",
+});
+
+/** Topic body sections worth carrying into the changelog (skipped while still `TODO`). */
+const NOTES = ["Trigger", "Invariant", "Trap"] as const;
+
+function bodyField(body: string, field: string): string | undefined {
+  return body.match(new RegExp(`^\\*\\*${field}:\\*\\*\\s*(.+)$`, "m"))?.[1]?.trim();
+}
+
+function bodySection(body: string, heading: string): string | undefined {
+  const m = body.match(new RegExp(`^### ${heading}\\n([\\s\\S]*?)(?=^### |(?![\\s\\S]))`, "m"));
+  const text = m?.[1]?.trim();
+  return text && text !== "TODO" ? text : undefined;
+}
+
+function render(f: Finished, tag: string, sha: string, files: Map<number, string[]>): string {
+  const { topic, problems } = f;
+  const finishDay = etDate(new Date(f.finishedAt));
+  // Snapshot at the finish instant: a problem closed later was open then.
+  const closedBy = (p: Problem) => p.closedAt !== null && p.closedAt <= f.finishedAt;
+  const onDay = problems.filter((p) => closedBy(p) && etDate(new Date(p.closedAt!)) === finishDay);
+  const earlier = problems.filter((p) => closedBy(p) && etDate(new Date(p.closedAt!)) !== finishDay);
+  const open = problems.filter((p) => !closedBy(p));
+
+  const row = (p: Problem): string[] => {
+    const lc = Number(p.title.match(/^LC (\d+)/)?.[1]);
+    const code = closedBy(p)
+      ? (files.get(lc) ?? []).map(
+          (path) => `[${path.slice(path.lastIndexOf(".") + 1)}](https://github.com/${gh.repo}/blob/${tag}/${encodeURI(path)})`,
+        )
+      : [];
+    return [
+      `#${p.number}`,
+      p.title,
+      labelValue(p, "set:") ?? "—",
+      labelValue(p, "diff:") ?? "—",
+      closedBy(p) ? ET_DAY.format(new Date(p.closedAt!)) : "—",
+      code.length ? [...code, `[docs](${DOCS_SITE}/${lc})`].join(" · ") : "—",
+    ];
+  };
+  const table = (list: Problem[]) =>
+    markdownTable([["Issue", "Problem", "Set", "Difficulty", "Closed (ET)", "Solution"], ...list.map(row)]);
+
+  // Blume's changelog derives each page's meta description from the opening
+  // prose, stripping markdown and line breaks without adding a space — so lead
+  // with one self-contained sentence and keep the fields on one `·` line.
+  const when = ET_WHEN.format(new Date(f.finishedAt));
+  const summary =
+    `Finished ${when} ET: all ${f.core.length} core problems closed, ` +
+    `${onDay.length} ${onDay.length === 1 ? "problem" : "problems"} closed that day.`;
+  const fields: string[] = [];
+  const scheduled = bodyField(topic.body, "Scheduled");
+  if (scheduled) fields.push(`**Scheduled:** ${scheduled}`);
+  if (topic.milestone) fields.push(`**Phase:** ${topic.milestone}`);
+  fields.push(`**Topic:** #${topic.number}`);
+  const buildsOn = bodyField(topic.body, "Builds on");
+  if (buildsOn) fields.push(`**Builds on:** ${buildsOn}`);
+  // The marker is how a later run finds this release; see the file header.
+  const sections = [`<!-- release-topics: #${topic.number} -->\n${summary}`, fields.join(" · ")];
+  sections.push(`### Closed ${ET_DAY.format(new Date(f.finishedAt))} — the day the topic finished\n\n${table(onDay)}`);
+  if (earlier.length) sections.push(`### Closed earlier\n\n${table(earlier)}`);
+  if (open.length) sections.push(`### Open at release\n\n${table(open)}`);
+  const notes = NOTES.flatMap((h) => {
+    const text = bodySection(topic.body, h);
+    return text ? [`**${h}.** ${text}`] : [];
+  });
+  if (notes.length) sections.push(`### Notes\n\n${notes.join("\n\n")}`);
+  sections.push(
+    `---\n\n\`${tag}\` is [timever](${TIMEVER}) for the UTC instant the core set completed; ` +
+      `it tags ${sha.slice(0, 7)}, the last main commit before it.`,
+  );
+  return `${sections.join("\n\n")}\n`;
+}
+
+// ── reconcile ────────────────────────────────────────────────────
+
+const latest = finished.at(-1)?.topic.number;
+const report: string[][] = [];
+const previews: string[] = [];
+let created = 0;
+let updated = 0;
+
+for (const f of finished) {
+  const { topic } = f;
+  const finishedEt = etDate(new Date(f.finishedAt));
+  const existing = released.get(topic.number);
+  // An existing release keeps its tag and commit; only its body is reconciled.
+  const tag = existing?.tag ?? tags.get(topic.number)!;
+  const sha = existing
+    ? git("rev-parse", `${tag}^{commit}`)
+    : git("rev-list", "-1", "--first-parent", `--before=${f.finishedAt}`, "HEAD");
+  if (!sha) {
+    report.push([`#${topic.number}`, topic.title, finishedEt, tag, "ERROR: no commit before finish"]);
+    continue;
+  }
+  // LC number -> its first-solve files in the tagged tree. Keyed by file name,
+  // like firstSolved(): tags before the bucket split see `work/<Difficulty>/…`,
+  // and the work/3 + work/7 re-solves are not the solve that closed the issue.
+  const files = new Map<number, string[]>();
+  for (const path of git("ls-tree", "-r", "--name-only", sha, "--", "work").split("\n")) {
+    const bucket = parseWorkPath(path.slice("work/".length))?.bucket;
+    const lc = path.match(/\/(\d+)\.[^/]+\.(?:js|py)$/)?.[1];
+    if (!lc || (bucket && bucket !== 1)) continue;
+    files.set(Number(lc), [...(files.get(Number(lc)) ?? []), path]);
+  }
+  const body = render(f, tag, sha, files);
+
+  if (existing) {
+    if (existing.body === body) {
+      report.push([`#${topic.number}`, topic.title, finishedEt, tag, "released"]);
+    } else if (DRY) {
+      previews.push(`## ${tag} — ${topic.title}\n\n${body}`);
+      report.push([`#${topic.number}`, topic.title, finishedEt, tag, "would update"]);
+    } else {
+      await gh.api(`/repos/${gh.repo}/releases/${existing.id}`, {
+        method: "PATCH",
+        body: JSON.stringify({ body }),
+      });
+      report.push([`#${topic.number}`, topic.title, finishedEt, tag, "updated"]);
+      updated++;
+    }
+    continue;
+  }
+
+  if (DRY) {
+    previews.push(`## ${tag} — ${topic.title}\n\n${body}`);
+    report.push([`#${topic.number}`, topic.title, finishedEt, tag, "would release"]);
+    continue;
+  }
+
+  const remote = git("ls-remote", "--tags", "origin", `refs/tags/${tag}`);
+  if (!remote) {
+    const local = Bun.spawnSync(["git", "rev-parse", "-q", "--verify", `refs/tags/${tag}`], { cwd: ROOT });
+    if (local.exitCode !== 0) {
+      const tagged = Bun.spawnSync(
+        ["git", "tag", "-a", tag, sha, "-m", `${topic.title} (#${topic.number})\n\nCore set complete: ${f.core.map((n) => `#${n}`).join(", ")}`],
+        { cwd: ROOT, env: { ...process.env, GIT_COMMITTER_DATE: f.finishedAt } },
+      );
+      if (tagged.exitCode !== 0) throw new Error(`git tag ${tag}: ${tagged.stderr.toString().trim()}`);
+    }
+    git("push", "origin", `refs/tags/${tag}`);
+  }
+  await gh.api(`/repos/${gh.repo}/releases`, {
+    method: "POST",
+    body: JSON.stringify({
+      tag_name: tag,
+      name: topic.title,
+      body,
+      make_latest: topic.number === latest ? "true" : "false",
+    }),
+  });
+  report.push([`#${topic.number}`, topic.title, finishedEt, tag, "released now"]);
+  created++;
+}
+
+// ── report ───────────────────────────────────────────────────────
+
+if (previews.length) console.log(`${previews.join("\n")}\n`);
+const rows = [["topic", "title", "finished (ET)", "tag", "status"], ...report];
+printTable(rows);
+const would = (status: string) => report.filter((r) => r[4] === status).length;
+console.log(
+  `\n${finished.length} finished topics · ` +
+    (DRY
+      ? `would release ${would("would release")}, update ${would("would update")}`
+      : `released ${created}, updated ${updated}`),
+);
+
+await writeStepSummary(
+  `### release-topics${DRY ? " (dry run)" : ""}\n\n` +
+    `${finished.length} finished topics, ${released.size} already released.\n\n` +
+    (report.length ? `${markdownTable(rows)}\n` : "No finished topics.\n") +
+    (previews.length ? `\n${previews.join("\n")}` : ""),
+);
@@ -30,17 +30,27 @@ function lcNumber(rel: string): number | undefined {
 }

 /** Fuzzy-pick a solution under work/ and run `leetcode <command>` on it. */
-export async function runOnSolution(command: "test" | "submit", message: string): Promise<never> {
+export async function runOnSolution(
+  command: "test" | "submit",
+  message: string,
+): Promise<never> {
  const db = problemDb();

-  const scanned = [...new Bun.Glob("**/*.{js,ts,py,java,c,cpp,go,rs,rb,swift,kt,cs}").scanSync({ cwd: WORK_DIR })];
+  const scanned = [
+    ...new Bun.Glob("**/*.{js,ts,py,java,c,cpp,go,rs,rb,swift,kt,cs}").scanSync(
+      { cwd: WORK_DIR },
+    ),
+  ];

  // register creation times so files older than the db still sort correctly
  for (const f of scanned) {
    const num = lcNumber(f);
    if (num === undefined) continue;
    const stat = statSync(join(WORK_DIR, f));
-    db.ensureSolution(num, new Date(stat.birthtimeMs || stat.mtimeMs).toISOString());
+    db.ensureSolution(
+      num,
+      new Date(stat.birthtimeMs || stat.mtimeMs).toISOString(),
+    );
  }
  const activity = db.solutionActivity();

@@ -51,20 +61,21 @@ export async function runOnSolution(command: "test" | "submit", message: string)
    return (command === "test" ? row?.lastTested : row?.createdAt) ?? undefined;
  };

-  const files = scanned
-    .sort((a, b) => {
-      const ka = sortKey(a);
-      const kb = sortKey(b);
-      if (ka !== kb) {
-        if (ka === undefined) return 1;
-        if (kb === undefined) return -1;
-        return kb.localeCompare(ka); // ISO timestamps: lexicographic = chronological
-      }
-      return a.localeCompare(b, undefined, { numeric: true });
-    });
+  const files = scanned.sort((a, b) => {
+    const ka = sortKey(a);
+    const kb = sortKey(b);
+    if (ka !== kb) {
+      if (ka === undefined) return 1;
+      if (kb === undefined) return -1;
+      return kb.localeCompare(ka); // ISO timestamps: lexicographic = chronological
+    }
+    return a.localeCompare(b, undefined, { numeric: true });
+  });

  if (files.length === 0) {
-    log.error(`No solution files found in ${relative(process.cwd(), WORK_DIR)}/`);
+    log.error(
+      `No solution files found in ${relative(process.cwd(), WORK_DIR)}/`,
+    );
    process.exit(1);
  }

@@ -74,11 +85,14 @@ export async function runOnSolution(command: "test" | "submit", message: string)
    maxItems: 12,
    options: files.map((f) => {
      const parsed = parseWorkPath(f);
-      const stage = parsed?.bucket === 1 ? "first solve" : `${parsed?.bucket}-day review`;
+      const stage =
+        parsed?.bucket === 1 ? "first solve" : `${parsed?.bucket}-day review`;
      return {
        value: f,
        label: basename(f),
-        hint: parsed ? `${stage} · ${parsed.difficulty} · ${parsed.category}` : f,
+        hint: parsed
+          ? `${stage} · ${parsed.difficulty} · ${parsed.category}`
+          : f,
      };
    }),
    filter: (search, option) => {
@@ -4,22 +4,44 @@
 *
 * - New problems get a full page (frontmatter, badge, warning, examples,
 *   constraints, solution) in the gold-standard format.
- * - Already-ported pages only get their `## Solution` section regenerated,
- *   so hand-curated prose is never touched.
+ * - Already-ported pages only get their script-owned parts regenerated — the
+ *   frontmatter `icon:` (`PAGE_ICON`), the `<Badge>` line and the
+ *   `## Solution` section — so hand-curated prose is never touched.
 * - When a problem is solved in both JavaScript and Python the solution
 *   renders as a <CodeGroup> (Python first); single-language solutions
 *   render as a plain fence.
 *
+ * Two classifications, two homes. The sidebar group is the campaign TOPIC
+ * from the README table (`tree-dfs/`, ordered by topic number via a
+ * script-owned `meta.ts`); the page badge is the data-structure CATEGORY the
+ * file sits under in `work/1/<Difficulty>/<Category>/`. A problem outside the
+ * plan (warm-ups) has no topic, so it groups under its category instead.
+ *
+ * Topics nest under their README PHASE: `(01-linear-structures)/tree-dfs/`.
+ * The phase folder is a Blume `(group)` — a sidebar section with its own
+ * script-owned `meta.ts`, adding no route segment — while the topic folder is
+ * a real segment (`/tree-dfs/invert-binary-tree`), so a hand-written
+ * `index.mdx` in it is the topic's landing page at `/tree-dfs` and Blume links
+ * the sidebar group row to it. Sync never writes or moves an `index.mdx`, and
+ * never removes a folder that still holds one. An unplanned category group has
+ * no phase and sits at the content root.
+ *
 * Only the first-solve bucket is read: work/3 and work/7 hold blind re-solves
 * of problems whose page already exists (see work.ts), and a fresh scaffold
 * there would otherwise wipe the page's `## Solution`.
+ *
+ * Every page also gets a short link: `/<num>` → `/<topic>/<slug>` (so
+ * `/leetcode/217` opens Contains Duplicate). The rules live in the generated
+ * `apps/docs/redirects.ts`, which `blume.config.ts` imports — Blume takes exact
+ * root-relative paths and prepends the deployment base itself.
 */
-import { mkdir, rename } from "node:fs/promises";
-import { basename, join } from "node:path";
+import { mkdir, readdir, rename, rmdir, unlink } from "node:fs/promises";
+import { basename, dirname, join } from "node:path";

 const ROOT = join(import.meta.dir, "..", "..");
 const WORK = join(ROOT, "work", "1");
 const DOCS = join(ROOT, "apps", "docs", "content");
+const REDIRECTS = join(ROOT, "apps", "docs", "redirects.ts");

 type Lang = "js" | "py";

@@ -158,24 +180,142 @@ function parseHeader(header: string): Header {

 // ── README topic map ─────────────────────────────────────────────

+/** A campaign phase: `## Phase N — Name · Weeks …` in the README. */
+interface Phase {
+  order: number;
+  name: string;
+}
+
+/** A campaign topic: `<strong>NN. Name</strong>` in the README table. */
+interface Topic {
+  order?: number; // README topic number; absent for an unplanned problem's category group
+  name: string;
+  phase?: Phase; // absent for an unplanned category group
+}
+
+/** README phase names → the short labels the overview page uses. */
+const PHASE_ALIASES: Record<string, string> = {
+  "Nodal & Grid Structures": "Nodal & Grid",
+  "Hierarchical Structures": "Hierarchical",
+  "Relational Structures": "Relational",
+  "Decision-Space Exploration": "Decision Space",
+};
+
+const ROMAN = ["I", "II", "III", "IV", "V", "VI", "VII", "VIII", "IX", "X"];
+
 const TOPIC_ALIASES: Record<string, string> = {
  "Hash-Based Lookup": "Hash Table",
 };

-async function readmeTopics(): Promise<Map<number, string>> {
+/**
+ * Sidebar icon per group, keyed by the (post-alias) topic name — the same key an
+ * unplanned problem's category group uses, so `Stack` the category and `Stack`
+ * the topic share one row. Lucide names; Blume resolves them at build time and
+ * reports a diagnostic for a typo, but an unmapped group is caught here first.
+ */
+const TOPIC_ICONS: Record<string, string> = {
+  // README topics, in curriculum order
+  "Hash Table": "hash",
+  "Prefix Sum": "sigma",
+  "Two Pointers": "arrow-left-right",
+  "Sliding Window": "app-window",
+  "Binary Search": "search",
+  Stack: "layers",
+  "Monotonic Stack": "trending-up",
+  "Linked List": "link",
+  "Hybrid Structures": "combine",
+  "Matrix Index Math": "grid-3x3",
+  "Tree DFS": "folder-tree",
+  "Tree BFS": "network",
+  "Binary Search Tree": "git-fork",
+  "Heap / Priority Queue": "triangle",
+  Trie: "spell-check",
+  "Graph DFS": "waypoints",
+  "Graph BFS": "radar",
+  "Topological Sort": "workflow",
+  "Union-Find": "merge",
+  Backtracking: "undo-2",
+  "1-D Dynamic Programming": "columns-3",
+  "Multi-D / Grid DP": "table",
+  Greedy: "coins",
+  Intervals: "calendar-range",
+  // work/1 categories that can surface as an unplanned group
+  Array: "brackets",
+  String: "type",
+  Tree: "network",
+  "Depth-First Search": "folder-tree",
+};
+
+/** LC number -> topic. A problem listed under two topics keeps the first. */
+async function readmeTopics(): Promise<Map<number, Topic>> {
  const md = await Bun.file(join(ROOT, "README.md")).text();
-  const map = new Map<number, string>();
-  let topic = "";
+  const map = new Map<number, Topic>();
+  let phase: Phase | undefined;
+  let topic: Topic | null = null;
  for (const line of md.split("\n")) {
-    const t = line.match(/<strong>\d+\.\s*([^<]+)<\/strong>/);
-    if (t) topic = TOPIC_ALIASES[t[1]!.trim()] ?? t[1]!.trim();
+    const p = line.match(/^## Phase ([IVX]+) — (.+?) · /);
+    if (p) {
+      const name = p[2]!.replaceAll("&amp;", "&");
+      phase = { order: ROMAN.indexOf(p[1]!) + 1, name: PHASE_ALIASES[name] ?? name };
+    }
+    const t = line.match(/<strong>(\d+)\.\s*([^<]+)<\/strong>/);
+    if (t) topic = { order: Number(t[1]), name: TOPIC_ALIASES[t[2]!.trim()] ?? t[2]!.trim(), phase };
    if (!topic) continue;
    const n = line.match(/<td align="center">(\d+)<\/td>/);
-    if (n) map.set(Number(n[1]), topic);
+    if (n && !map.has(Number(n[1]))) map.set(Number(n[1]), topic);
  }
  return map;
 }

+/**
+ * Folder slug for a topic or phase name. Blume reads a leading `<digits>-` on a
+ * folder as a sidebar-order prefix and strips it from the route (`1-d-…` →
+ * `/d-…`), so a slug that would start that way keeps the digits glued to the
+ * next token (`1d-dynamic-programming`).
+ */
+function slugify(name: string): string {
+  return name
+    .toLowerCase()
+    .replace(/[^a-z0-9]+/g, "-")
+    .replace(/^-|-$/g, "")
+    .replace(/^(\d+)-/, "$1");
+}
+
+/**
+ * Content folder for a topic, relative to `DOCS`: `(0N-<phase>)/<topic>` for a
+ * planned topic (the phase is a `(group)`, so the route is still `/<topic>`),
+ * bare `<topic>` for an unplanned category group.
+ */
+function topicDir(topic: Topic): string {
+  const slug = slugify(topic.name);
+  return topic.phase ? `${phaseDir(topic.phase)}/${slug}` : slug;
+}
+
+function phaseDir(phase: Phase): string {
+  return `(${String(phase.order).padStart(2, "0")}-${slugify(phase.name)})`;
+}
+
+/** Blume `(group)` meta for a phase: the overview page's heading, in phase order. */
+function phaseMeta(phase: Phase): string {
+  return `import { defineMeta } from "blume";\n\nexport default defineMeta({ title: ${JSON.stringify(`Phase ${phase.order}: ${phase.name}`)}, order: ${phase.order} });\n`;
+}
+
+/**
+ * Blume folder meta for a topic: exact label (the folder slug loses `/` and
+ * case), curriculum order, the sidebar icon, and `display: "group"` so the
+ * topic collapses under its phase heading. An unplanned category group
+ * (warm-ups) has no README number and sorts at 0, ahead of phase 1, so the
+ * sidebar opens on the basics.
+ */
+function groupMeta(topic: Topic): string {
+  const icon = TOPIC_ICONS[topic.name];
+  if (!icon) console.warn(`no icon for group ${JSON.stringify(topic.name)}: add it to TOPIC_ICONS`);
+  const fields = [`title: ${JSON.stringify(topic.name)}`, `order: ${topic.order ?? 0}`];
+  if (icon) fields.push(`icon: ${JSON.stringify(icon)}`);
+  fields.push(`display: "group"`);
+  return `import { defineMeta } from "blume";\n\nexport default defineMeta({ ${fields.join(", ")} });\n`;
+}
+
 // ── inline-code formatting heuristics ────────────────────────────

 /** Identifiers worth backticking in prose, harvested from the code. */
@@ -277,7 +417,14 @@ function warningText(h: Header): string {
  );
 }

-function renderPage(sol: Solution, topic: string): string {
+/** Sidebar/page icon for every problem page (Lucide, resolved by Blume). */
+const PAGE_ICON = "dot";
+
+function badgeLine(sol: Solution): string {
+  return `<Badge variant="accent">${sol.category}</Badge>`;
+}
+
+function renderPage(sol: Solution): string {
  const h = sol.header;
  const ids = identifiers(sol);
  const description = h.statement[0]?.replace(/\.\s*$/, "") ?? "";
@@ -287,12 +434,13 @@ function renderPage(sol: Solution, topic: string): string {
    "---",
    `title: '${h.num}. ${h.title.replaceAll("'", "''")}'`,
    /[:#]/.test(description) ? `description: '${description.replaceAll("'", "''")}'` : `description: ${description}`,
+    `icon: ${PAGE_ICON}`,
    "sidebar:",
    `  label: '${h.title.replaceAll("'", "''")}'`,
    `  badge: '${h.difficulty}'`,
    "---",
    "",
-    `<Badge variant="accent">${topic}</Badge>`,
+    badgeLine(sol),
    "",
  );

@@ -350,6 +498,11 @@ for await (const rel of new Bun.Glob("**/*.{js,py}").scan(WORK)) {
 }

 const topics = await readmeTopics();
+// An unplanned problem whose work/ category is also a README topic name (a
+// `Hash Table` warm-up) files under that topic, in its phase, not a second
+// root-level `hash-table/`.
+const topicByName = new Map<string, Topic>();
+for (const topic of topics.values()) topicByName.set(topic.name, topic);

 /** Insert a number-free `sidebar.label` into existing frontmatter when missing. */
 function ensureLabel(page: string, h: Header): string {
@@ -359,6 +512,17 @@ function ensureLabel(page: string, h: Header): string {
  return page.replace(/^sidebar:$/m, `sidebar:\n${label}`);
 }

+/** Set the frontmatter's top-level `icon:` to `PAGE_ICON`, inserting it before `sidebar:`. */
+function ensureIcon(page: string): string {
+  const fm = page.match(/^---\n[\s\S]*?\n---/);
+  if (!fm) return page;
+  const line = `icon: ${PAGE_ICON}`;
+  const next = /^icon: .*$/m.test(fm[0])
+    ? fm[0].replace(/^icon: .*$/m, line)
+    : fm[0].replace(/^sidebar:$/m, `${line}\nsidebar:`);
+  return next + page.slice(fm[0].length);
+}
+
 /**
 * Offset just past the script-owned `## Solution` section: the next top-level
 * heading that is not inside a code fence, or end of file. Sections a human
@@ -387,15 +551,24 @@ for (const rel of (await Array.fromAsync(new Bun.Glob("**/*.mdx").scan(DOCS))).s

 const report: [string, string, string][] = [];
 const ordered = [...solutions.values()].sort((a, b) => a.num - b.num);
+const groups = new Map<string, Topic>(); // topic folder -> topic, for meta.ts
+const phases = new Map<string, Phase>(); // phase folder -> phase, for meta.ts
+const vacated = new Set<string>(); // folders a page moved out of; removed below if empty
+const redirects: string[] = []; // `{ from: "/<num>", to: "/<topic>/<slug>" }` rows, numeric order

 for (const sol of ordered) {
  const langs = Object.keys(sol.code).sort().join("+");
  const workRef = `${sol.num}.${sol.slug} (${langs})`;
-  // (category) group folder + leetcode-number prefix: numeric sidebar order, both stripped from the URL.
-  const group = `(${sol.category.toLowerCase().replaceAll(" ", "-")})`;
-  const name = `${group}/${sol.num}-${sol.slug}.mdx`;
+  // (0N-<phase>)/<topic>/ folder (the topic is a real route segment, so its
+  // index.mdx is the landing page) + leetcode-number prefix: numeric sidebar
+  // order, prefix stripped from the URL.
+  const topic = topics.get(sol.num) ?? topicByName.get(sol.category) ?? { name: sol.category };
+  const dir = topicDir(topic);
+  groups.set(dir, topic);
+  if (topic.phase) phases.set(phaseDir(topic.phase), topic.phase);
+  const name = `${dir}/${sol.num}-${sol.slug}.mdx`;
  const existing = pages.get(sol.num);
-  await mkdir(join(DOCS, group), { recursive: true });
+  await mkdir(join(DOCS, dir), { recursive: true });

  if (existing) {
    const page = await Bun.file(join(DOCS, existing)).text();
@@ -406,10 +579,11 @@ for (const sol of ordered) {
    }
    const tail = page.slice(solutionEnd(page, at));
    const body = page.slice(0, at) + solutionSection(sol) + (tail ? `\n${tail}` : "");
-    const next = ensureLabel(body, sol.header);
+    const next = ensureIcon(ensureLabel(body, sol.header)).replace(/^<Badge variant="accent">.*<\/Badge>$/m, badgeLine(sol));
    const actions: string[] = [];
    if (existing !== name) {
      await rename(join(DOCS, existing), join(DOCS, name));
+      vacated.add(dirname(existing));
      actions.push("moved");
    }
    if (next !== page) {
@@ -418,10 +592,59 @@ for (const sol of ordered) {
    }
    report.push([workRef, name, actions.join(" + ") || "skipped (in sync)"]);
  } else {
-    const topic = topics.get(sol.num) ?? sol.category;
-    await Bun.write(join(DOCS, name), renderPage(sol, topic));
+    await Bun.write(join(DOCS, name), renderPage(sol));
    report.push([workRef, name, "created"]);
  }
+  // The route drops the phase `(group)` and the page's number prefix.
+  redirects.push(`  { from: "/${sol.num}", to: ${JSON.stringify(`/${slugify(topic.name)}/${sol.slug}`)} },`);
+}
+
+// Script-owned short links, one per page; imported by blume.config.ts.
+{
+  const module = [
+    "// Generated by `bun run sync` (apps/cli/sync.ts) — do not edit.",
+    "// `/leetcode/<num>` → that problem's page; exact paths, base added by Blume.",
+    "export default [",
+    ...redirects,
+    "];",
+    "",
+  ].join("\n");
+  if (!(await Bun.file(REDIRECTS).exists()) || (await Bun.file(REDIRECTS).text()) !== module) {
+    await Bun.write(REDIRECTS, module);
+    report.push([`${redirects.length} redirects`, "apps/docs/redirects.ts", "written"]);
+  }
+}
+
+// Script-owned folder metas — one per live phase and topic folder — rewritten
+// only when they differ.
+const metas: [label: string, dir: string, meta: string][] = [
+  ...[...phases].map(([dir, phase]): [string, string, string] => [`phase ${phase.order}`, dir, phaseMeta(phase)]),
+  ...[...groups].map(([dir, topic]): [string, string, string] => [`topic ${topic.order ?? "-"}`, dir, groupMeta(topic)]),
+];
+for (const [label, dir, meta] of metas) {
+  const path = join(DOCS, dir, "meta.ts");
+  if ((await Bun.file(path).exists()) && (await Bun.file(path).text()) === meta) continue;
+  await Bun.write(path, meta);
+  report.push([label, `${dir}/meta.ts`, "written"]);
+}
+
+// A folder every page left behind is dead: drop its meta and the folder itself,
+// then the phase folder above it if that emptied too. A folder that still holds
+// anything else (a hand-written index.mdx, another topic) stops the climb.
+for (const start of vacated) {
+  for (let dir = start; dir !== "."; dir = dirname(dir)) {
+    if (groups.has(dir) || phases.has(dir)) break;
+    let left: string[];
+    try {
+      left = await readdir(join(DOCS, dir));
+    } catch {
+      break; // already removed via a sibling's climb
+    }
+    if (left.some((f) => f !== "meta.ts")) break;
+    if (left.length) await unlink(join(DOCS, dir, "meta.ts"));
+    await rmdir(join(DOCS, dir));
+    report.push(["", dir, "removed (empty)"]);
+  }
 }

 const w0 = Math.max(...report.map((r) => r[0].length), 9);
@@ -0,0 +1,75 @@
+/**
+ * Topic issues and their problem sub-issues: the ONE reading of "is this topic
+ * finished", shared by close-topics.ts (closes the topic issue) and
+ * release-topics.ts (publishes the topic's timever release) so the two can
+ * never disagree about when a topic was done.
+ *
+ * Rule: a topic is finished when it has a core set (`set:core` sub-issues) and
+ * every one of them is closed. Optional and deferred problems are extra credit
+ * and never hold a topic open. A topic with no core sub-issues is never
+ * finished — that would be indistinguishable from a mis-labelled problem set.
+ *
+ * Library: side-effect-free on import; every call takes the caller's github().
+ */
+import type { GitHub } from "./github.ts";
+
+/** One problem sub-issue, narrowed from the sub_issues payload. */
+export interface Problem {
+  number: number;
+  /** `LC <num> · <Name>`. */
+  title: string;
+  /** ISO instant the issue closed; null while it is open. */
+  closedAt: string | null;
+  labels: string[];
+}
+
+export interface Core {
+  /** Core sub-issues, closed and open alike, lowest number first. */
+  all: number[];
+  /** The core sub-issues still open — non-empty means the topic stays open. */
+  pending: number[];
+  /** When the last core sub-issue closed; null until the topic is finished. */
+  finishedAt: string | null;
+}
+
+/**
+ * Every sub-issue of a topic. The sub_issues payload carries the full issue
+ * objects (state + labels), so no per-problem follow-up request is needed.
+ */
+export async function readProblems(gh: GitHub, topic: number): Promise<Problem[]> {
+  const problems: Problem[] = [];
+  for await (const raw of gh.list(`/repos/${gh.repo}/issues/${topic}/sub_issues`)) {
+    if (!raw || typeof raw !== "object") continue;
+    if (!("number" in raw) || typeof raw.number !== "number") continue;
+    if (!("title" in raw) || typeof raw.title !== "string") continue;
+    if (!("state" in raw) || typeof raw.state !== "string") continue;
+    if (!("labels" in raw) || !Array.isArray(raw.labels)) continue;
+    const closedAt =
+      raw.state === "closed" && "closed_at" in raw && typeof raw.closed_at === "string"
+        ? raw.closed_at
+        : null;
+    const labels = raw.labels.flatMap((label: unknown) =>
+      label && typeof label === "object" && "name" in label && typeof label.name === "string"
+        ? [label.name]
+        : [],
+    );
+    problems.push({ number: raw.number, title: raw.title, closedAt, labels });
+  }
+  return problems.sort((a, b) => a.number - b.number);
+}
+
+/** The value of a `prefix:value` label (`set:` → `core`), if the problem has one. */
+export function labelValue(problem: Problem, prefix: string): string | undefined {
+  return problem.labels.find((l) => l.startsWith(prefix))?.slice(prefix.length);
+}
+
+export function coreSet(problems: Problem[]): Core {
+  const core = problems.filter((p) => labelValue(p, "set:") === "core");
+  const pending = core.filter((p) => !p.closedAt).map((p) => p.number);
+  // ISO instants from the API share one format, so string order is time order.
+  const finishedAt =
+    core.length && !pending.length
+      ? core.map((p) => p.closedAt!).reduce((a, b) => (a > b ? a : b))
+      : null;
+  return { all: core.map((p) => p.number), pending, finishedAt };
+}
@@ -0,0 +1,173 @@
+#!/usr/bin/env bun
+/**
+ * Find the NeetCode explanation video for every docs page that has none.
+ *
+ * A finished solution page ends with `## Explanation` holding one
+ * `<YouTube id="…" title="…" />` — NeetCode's walkthrough of the problem.
+ * `sync.ts` never writes that section (it owns only `## Solution`), so pages
+ * accumulate without one. This script lists every numbered page lacking a
+ * `<YouTube` embed and looks each up on NeetCode's two channels, main channel
+ * first, accepting only a title that names the page's exact LC number
+ * (`… - Leetcode 226 - …`) and is not a #short. No YouTube API key: the
+ * channel search page embeds its results as `ytInitialData` JSON, which is
+ * all we parse. Ranking is the channel's own — the first exact-number match
+ * wins.
+ *
+ *   bun apps/cli/videos.ts           # report: page → video (or "no match")
+ *   bun apps/cli/videos.ts --write   # also append `## Explanation` + embed
+ *
+ * --write appends at end of file (below `## Solution`, where sync.ts leaves
+ * human sections alone) and adds the `## Explanation` heading only when the
+ * page lacks it. A page with no match is left untouched, so re-runs are
+ * no-ops once every page has its embed.
+ */
+import { join } from "node:path";
+
+import { printTable } from "./report.ts";
+import { ROOT } from "./work.ts";
+
+const DOCS = join(ROOT, "apps", "docs", "content");
+const WRITE = process.argv.includes("--write");
+
+/** Channel handles searched in order; the first exact match wins. */
+const CHANNELS = ["NeetCode", "NeetCodeIO"] as const;
+
+interface Video {
+  id: string;
+  title: string;
+  channel: string;
+}
+
+// ── YouTube channel search (no API key) ──────────────────────────
+
+const HEADERS = { "accept-language": "en-US,en;q=0.9", "user-agent": "Mozilla/5.0" };
+
+/**
+ * The `var ytInitialData = {…};</script>` blob of a YouTube page, parsed.
+ * YouTube intermittently answers 200 with a page that lacks the blob or cuts
+ * it short, so a bad page is retried a few times before it is an error.
+ */
+async function initialData(url: string): Promise<unknown> {
+  let reason = "";
+  for (let attempt = 0; attempt < 3; attempt++) {
+    if (attempt) await Bun.sleep(500 * attempt);
+    const res = await fetch(url, { headers: HEADERS });
+    if (!res.ok) {
+      reason = `status ${res.status}`;
+      continue;
+    }
+    const html = await res.text();
+    // Slice rather than regex: the blob is ~1 MB, and a lazy `.*?` across it
+    // is what made the first cut of this script fail at random.
+    const open = html.search(/ytInitialData\s*=\s*\{/);
+    const close = open === -1 ? -1 : html.indexOf("</script>", open);
+    if (close === -1) {
+      reason = `no ytInitialData in ${html.length}-byte page`;
+      continue;
+    }
+    try {
+      return JSON.parse(html.slice(html.indexOf("{", open), close).replace(/;\s*$/, ""));
+    } catch {
+      reason = "truncated ytInitialData";
+    }
+  }
+  throw new Error(`GET ${url} -> ${reason}`);
+}
+
+/** Every videoRenderer in a channel search page, in the page's own order. */
+async function channelSearch(channel: string, query: string): Promise<Video[]> {
+  const url = `https://www.youtube.com/@${channel}/search?query=${encodeURIComponent(query)}`;
+  const data = await initialData(url);
+
+  const videos: Video[] = [];
+  const walk = (node: unknown): void => {
+    if (!node || typeof node !== "object") return;
+    if ("videoRenderer" in node) {
+      const v = node.videoRenderer;
+      if (v && typeof v === "object" && "videoId" in v && typeof v.videoId === "string") {
+        const runs =
+          "title" in v && v.title && typeof v.title === "object" && "runs" in v.title
+            ? v.title.runs
+            : undefined;
+        const title = Array.isArray(runs)
+          ? runs.map((r) => (r && typeof r === "object" && "text" in r ? String(r.text) : "")).join("")
+          : "";
+        videos.push({ id: v.videoId, title, channel });
+      }
+      return;
+    }
+    for (const child of Object.values(node)) walk(child);
+  };
+  walk(data);
+  return videos;
+}
+
+/** The LC number a NeetCode title names (`… - Leetcode 226 - …`), if any. */
+function titledNumber(title: string): number | undefined {
+  const m = title.match(/\bleetcode\s*#?\s*(\d+)\b/i);
+  return m ? Number(m[1]) : undefined;
+}
+
+async function findVideo(num: number): Promise<Video | undefined> {
+  for (const channel of CHANNELS) {
+    const hit = (await channelSearch(channel, `leetcode ${num}`)).find(
+      (v) => titledNumber(v.title) === num && !/#shorts\b/i.test(v.title),
+    );
+    if (hit) return hit;
+  }
+  return undefined;
+}
+
+// ── docs pages ───────────────────────────────────────────────────
+
+interface Page {
+  num: number;
+  rel: string;
+  text: string;
+}
+
+/** Numbered solution pages without a `<YouTube` embed, by LC number. */
+async function pagesMissingVideo(): Promise<Page[]> {
+  const pages: Page[] = [];
+  for (const rel of await Array.fromAsync(new Bun.Glob("**/*.mdx").scan(DOCS))) {
+    const text = await Bun.file(join(DOCS, rel)).text();
+    const t = text.match(/^title: '(\d+)\./m);
+    if (t && !text.includes("<YouTube")) pages.push({ num: Number(t[1]), rel, text });
+  }
+  return pages.sort((a, b) => a.num - b.num);
+}
+
+function withEmbed(page: Page, video: Video): string {
+  const embed = `<YouTube id="${video.id}" title="${video.title.replaceAll('"', "'")}" />`;
+  const heading = /^## Explanation\s*$/m.test(page.text) ? "" : "## Explanation\n\n";
+  return `${page.text.replace(/\s*$/, "\n")}\n${heading}${embed}\n`;
+}
+
+// ── main ─────────────────────────────────────────────────────────
+
+const missing = await pagesMissingVideo();
+if (missing.length === 0) {
+  console.log("every solution page has a video");
+  process.exit(0);
+}
+
+const rows: string[][] = [["docs page", "video", "title", WRITE ? "status" : ""]];
+let found = 0;
+for (const page of missing) {
+  const video = await findVideo(page.num);
+  if (!video) {
+    rows.push([page.rel, "—", "no NeetCode match", WRITE ? "skipped" : ""]);
+    continue;
+  }
+  found++;
+  let status = "";
+  if (WRITE) {
+    await Bun.write(join(DOCS, page.rel), withEmbed(page, video));
+    status = "written";
+  }
+  rows.push([page.rel, `https://youtu.be/${video.id}`, `${video.title} (${video.channel})`, status]);
+}
+
+if (!WRITE) for (const row of rows) row.pop();
+printTable(rows);
+console.log(`\n${found}/${missing.length} page(s) matched${WRITE ? `, ${found} written` : ""}`);
@@ -1,21 +1,64 @@
 import { defineConfig } from "blume";
+import { filesystem, githubReleases } from "blume/sources";
+import redirects from "./redirects";
+
+// Blume's defaults (`inter`, `ibm-plex-mono`) are fetched from fonts.gstatic.com
+// on every build, which has timed out on the Pages runner and, fully offline,
+// silently ships no fonts at all. The same two families from Fontsource, as
+// local files, keep the build network-free. Blume resolves `src` from this
+// directory; the workspace links these packages under ./node_modules.
+const inter = {
+  name: "Inter",
+  variants: [
+    {
+      src: "./node_modules/@fontsource-variable/inter/files/inter-latin-wght-normal.woff2",
+      weight: "100..900",
+    },
+  ],
+};
+const plexMono = {
+  name: "IBM Plex Mono",
+  variants: [
+    {
+      src: "./node_modules/@fontsource/ibm-plex-mono/files/ibm-plex-mono-latin-400-normal.woff2",
+      weight: 400,
+    },
+    {
+      src: "./node_modules/@fontsource/ibm-plex-mono/files/ibm-plex-mono-latin-500-normal.woff2",
+      weight: 500,
+    },
+  ],
+};

 export default defineConfig({
-  title: "Leetcode",
-  description: "Documentation for LeetCode solutions",
+  title: "prdlk",
+  description: "Documentation for my LeetCode solutions",
+  logo: "./public/logo.svg",
  deployment: {
-    site: "https://prdlk.github.io",
-    base: "/leetcode",
+    site: "https://lc.prd.lk",
  },
+  github: {
+    owner: "prdlk",
+    repo: "leetcode",
+  },
+  lastModified: "git",
+  redirects,
  content: {
-    root: "content",
-  },
-  navigation: {
-    tabs: [
-      {
-        label: "Reference",
-        path: "/reference",
-      },
+    sources: [
+      filesystem({ root: "content" }),
+      githubReleases({
+        prefix: "changelog",
+        owner: "prdlk",
+        repo: "leetcode",
+      }),
    ],
  },
+  theme: {
+    fonts: { display: inter, body: inter, mono: plexMono },
+  },
+  navigation: {
+    sidebar: {
+      display: "flat", // "flat" | "group" | "page"
+    },
+  },
 });
@@ -0,0 +1,13 @@
+/**
+ * Project MDX components, available on every page with no import. Static
+ * `.astro` files ship no framework JS (a custom element carries any
+ * behaviour); interactive React lives in `islands/` instead.
+ */
+import { defineComponents } from "blume";
+
+export default defineComponents({
+  mdx: {
+    Carousel: "./components/Carousel.astro",
+    CarouselImage: "./components/CarouselImage.astro",
+  },
+});
@@ -0,0 +1,170 @@
+---
+/**
+ * A horizontal carousel of <CarouselImage> slides, built the way Blume's own
+ * content components are: static markup, a custom element for the behaviour,
+ * no framework. The track is a CSS scroll-snap strip (so it also swipes on
+ * touch and the browser does the animation); prev/next sit outside it and
+ * the play toggle sits below with a "n / N" counter.
+ *
+ * `time` is seconds per slide while playing (default 1.25); `autoplay` starts
+ * playing on load. Play wraps around, pauses whenever the tab is hidden, and
+ * never starts on its own for a reader who prefers reduced motion — the
+ * button still works for them. Manual navigation (buttons, arrow keys, swipe)
+ * does not stop playback; it only restarts the interval so a step is never
+ * doubled.
+ */
+import Icon from "blume/components/Icon.astro";
+
+interface Props {
+  /** Seconds each slide stays while playing. */
+  time?: number;
+  autoplay?: boolean;
+}
+
+const { time = 1.25, autoplay = false } = Astro.props;
+
+const buttonClass =
+  "inline-flex size-9 shrink-0 cursor-pointer items-center justify-center rounded-full border border-border bg-background text-muted-foreground transition-colors hover:border-foreground hover:text-foreground disabled:cursor-default disabled:opacity-40";
+---
+
+<blume-carousel
+  class="not-prose my-6 block"
+  data-time={String(time)}
+  data-autoplay={autoplay ? "" : undefined}
+  role="region"
+  aria-roledescription="carousel"
+>
+  <div class="flex items-center gap-3">
+    <button class={buttonClass} data-prev type="button" aria-label="Previous slide">
+      <Icon class="rtl:-scale-x-100" name="chevron-left" size={18} />
+    </button>
+    <div
+      class="flex min-w-0 flex-1 snap-x snap-mandatory overflow-x-auto scroll-smooth rounded-blume [scrollbar-width:none] [&::-webkit-scrollbar]:hidden"
+      data-track
+      tabindex="0"
+      aria-live="polite"
+    >
+      <slot />
+    </div>
+    <button class={buttonClass} data-next type="button" aria-label="Next slide">
+      <Icon class="rtl:-scale-x-100" name="chevron-right" size={18} />
+    </button>
+  </div>
+  <div class="mt-3 flex items-center justify-center gap-3">
+    <button class={buttonClass} data-play type="button" aria-label="Play" aria-pressed="false">
+      <Icon class="[[aria-pressed=true]_&]:hidden" name="play" size={16} />
+      <Icon class="[[aria-pressed=false]_&]:hidden" name="pause" size={16} />
+    </button>
+    <span class="text-muted-foreground text-xs tabular-nums leading-5" data-count></span>
+  </div>
+</blume-carousel>
+
+<script>
+  class BlumeCarousel extends HTMLElement {
+    #track!: HTMLElement;
+    #play!: HTMLButtonElement;
+    #count!: HTMLElement;
+    #timer: number | null = null;
+    #index = 0;
+
+    connectedCallback() {
+      this.#track = this.querySelector("[data-track]")!;
+      this.#play = this.querySelector("[data-play]")!;
+      this.#count = this.querySelector("[data-count]")!;
+      const prev = this.querySelector<HTMLButtonElement>("[data-prev]")!;
+      const next = this.querySelector<HTMLButtonElement>("[data-next]")!;
+
+      const slides = this.#slides().length;
+      if (slides < 2) {
+        for (const b of [prev, next, this.#play]) b.disabled = true;
+      }
+      this.#render();
+
+      prev.addEventListener("click", () => this.#step(-1));
+      next.addEventListener("click", () => this.#step(1));
+      this.#play.addEventListener("click", () => (this.#timer === null ? this.#start() : this.#stop()));
+      this.addEventListener("keydown", (event) => {
+        if (event.key === "ArrowLeft") this.#step(-1);
+        else if (event.key === "ArrowRight") this.#step(1);
+        else return;
+        event.preventDefault();
+      });
+      // Swipes and trackpad scrolls move the track without #step; follow them
+      // once the scroll settles. Settle-only, not per event: a long smooth
+      // scroll from #step passes over intermediate slides and must not be
+      // read back as the reader's position mid-flight.
+      let settle: number | undefined;
+      this.#track.addEventListener(
+        "scroll",
+        () => {
+          window.clearTimeout(settle);
+          settle = window.setTimeout(() => this.#syncFromScroll(), 150);
+        },
+        { passive: true }
+      );
+      document.addEventListener("visibilitychange", this.#onVisibility);
+
+      const reduced = matchMedia("(prefers-reduced-motion: reduce)").matches;
+      if (this.hasAttribute("data-autoplay") && slides > 1 && !reduced) this.#start();
+    }
+
+    disconnectedCallback() {
+      this.#stop();
+      document.removeEventListener("visibilitychange", this.#onVisibility);
+    }
+
+    #slides(): HTMLElement[] {
+      return [...this.#track.querySelectorAll<HTMLElement>(":scope > blume-carousel-image")];
+    }
+
+    #step(delta: number) {
+      const slides = this.#slides();
+      this.#index = (this.#index + delta + slides.length) % slides.length;
+      slides[this.#index]!.scrollIntoView({ block: "nearest", inline: "start", behavior: "smooth" });
+      this.#render();
+      if (this.#timer !== null) this.#start(); // restart the interval from this slide
+    }
+
+    #syncFromScroll() {
+      const at = Math.round(this.#track.scrollLeft / this.#track.clientWidth);
+      if (at !== this.#index && Number.isFinite(at)) {
+        this.#index = Math.min(Math.max(at, 0), this.#slides().length - 1);
+        this.#render();
+      }
+    }
+
+    #start() {
+      this.#stop();
+      const ms = Number(this.dataset.time) * 1000;
+      this.#timer = window.setInterval(() => this.#step(1), ms > 0 ? ms : 1250);
+      this.#play.setAttribute("aria-pressed", "true");
+      this.#play.setAttribute("aria-label", "Pause");
+    }
+
+    #stop() {
+      if (this.#timer !== null) window.clearInterval(this.#timer);
+      this.#timer = null;
+      this.#play.setAttribute("aria-pressed", "false");
+      this.#play.setAttribute("aria-label", "Play");
+    }
+
+    // A background tab keeps its timers but not its layout; resume on return.
+    #wasPlaying = false;
+    #onVisibility = () => {
+      if (document.hidden) {
+        this.#wasPlaying = this.#timer !== null;
+        this.#stop();
+      } else if (this.#wasPlaying) this.#start();
+    };
+
+    #render() {
+      const slides = this.#slides();
+      this.#count.textContent = `${this.#index + 1} / ${slides.length}`;
+      slides.forEach((s, i) => s.setAttribute("aria-hidden", i === this.#index ? "false" : "true"));
+    }
+  }
+
+  if (!customElements.get("blume-carousel")) {
+    customElements.define("blume-carousel", BlumeCarousel);
+  }
+</script>
@@ -0,0 +1,45 @@
+---
+/**
+ * One slide of a <Carousel>: Blume's own <Frame> around an <img>, so a slide
+ * is pixel-identical to a framed image elsewhere on the site — `alt` doubles
+ * as the frame's hint line and `caption` is Markdown, exactly as in Frame.
+ *
+ * `src` accepts what MDX authors already write for images: a URL, a
+ * root-relative `/file.png` (served from public/, prefixed with the deploy
+ * base), or the page-relative `../../public/file.png` form. A relative path
+ * that does not pass through public/ has nothing to serve it, so it is a
+ * build error rather than a silent 404.
+ */
+import Frame from "blume/components/content/Frame.astro";
+import { withBase } from "blume/components/islands/base-path.ts";
+
+interface Props {
+  src: string;
+  alt?: string;
+  caption?: string;
+}
+
+const { src, alt, caption } = Astro.props;
+
+const resolveSrc = (value: string): string => {
+  if (URL.canParse(value)) return value;
+  if (value.startsWith("/")) return withBase(value);
+  const at = value.indexOf("public/");
+  if (at === -1) {
+    throw new Error(
+      `<CarouselImage src="${value}">: a relative src must point into public/ (or use /file.png)`
+    );
+  }
+  return withBase(`/${value.slice(at + "public/".length)}`);
+};
+---
+
+<blume-carousel-image
+  class="block w-full shrink-0 snap-center [&_blume-frame]:my-0"
+  role="group"
+  aria-roledescription="slide"
+>
+  <Frame caption={caption} hint={alt}>
+    <img alt={alt ?? ""} src={resolveSrc(src)} loading="lazy" draggable="false" />
+  </Frame>
+</blume-carousel-image>
@@ -1,12 +1,13 @@
 ---
 title: '153. Find Minimum in Rotated Sorted Array'
 description: 'Suppose an array of length n sorted in ascending order is rotated between 1 and n times. For example, the array nums = [0,1,2,4,5,6,7] might become:'
+icon: dot
 sidebar:
  label: 'Find Minimum in Rotated Sorted Array'
  badge: 'Medium'
 ---

-<Badge variant="accent">Binary Search</Badge>
+<Badge variant="accent">Array</Badge>

 ::::warning
 Notice that rotating an array [a[0], a[1], a[2], ..., a[n-1]] 1 time results in the array [a[n-1], a[0], a[1], a[2], ..., a[n-2]].
@@ -1,12 +1,13 @@
 ---
 title: '33. Search in Rotated Sorted Array'
 description: There is an integer array nums sorted in ascending order (with distinct values)
+icon: dot
 sidebar:
  label: 'Search in Rotated Sorted Array'
  badge: 'Medium'
 ---

-<Badge variant="accent">Binary Search</Badge>
+<Badge variant="accent">Array</Badge>

 ::::warning
 You must write an algorithm with O(log n) runtime complexity.
@@ -1,12 +1,13 @@
 ---
 title: '35. Search Insert Position'
 description: Given a sorted array of distinct integers and a target value, return the index if the target is found. If not, return the index where it would be if it were inserted in order
+icon: dot
 sidebar:
  label: 'Search Insert Position'
  badge: 'Easy'
 ---

-<Badge variant="accent">Binary Search</Badge>
+<Badge variant="accent">Array</Badge>

 ::::warning
 You must write an algorithm with O(log n) runtime complexity.
@@ -1,12 +1,13 @@
 ---
 title: '704. Binary Search'
 description: Given an array of integers nums which is sorted in ascending order, and an integer target, write a function to search target in nums. If target exists, then return its index. Otherwise, return -1
+icon: dot
 sidebar:
  label: 'Binary Search'
  badge: 'Easy'
 ---

-<Badge variant="accent">Binary Search</Badge>
+<Badge variant="accent">Array</Badge>

 ::::warning
 You must write an algorithm with O(log n) runtime complexity.
@@ -1,12 +1,13 @@
 ---
 title: '74. Search a 2D Matrix'
 description: 'You are given an m x n integer matrix matrix with the following two properties:'
+icon: dot
 sidebar:
  label: 'Search a 2D Matrix'
  badge: 'Medium'
 ---

-<Badge variant="accent">Binary Search</Badge>
+<Badge variant="accent">Array</Badge>

 ::::warning
 You must write a solution in O(log(m * n)) time complexity.
@@ -1,12 +1,13 @@
 ---
 title: '875. Koko Eating Bananas'
 description: Koko loves to eat bananas. There are n piles of bananas, the i^th pile has piles[i] bananas. The guards have gone and will come back in h hours
+icon: dot
 sidebar:
  label: 'Koko Eating Bananas'
  badge: 'Medium'
 ---

-<Badge variant="accent">Binary Search</Badge>
+<Badge variant="accent">Array</Badge>

 ### Example 1:
 - Input: `piles = [3,6,7,11], h = 8`
@@ -0,0 +1,5 @@
+---
+title: Binary Search
+description: 'Half of a sorted range is removed at each step until the target is found.'
+hidden: true
+---
@@ -0,0 +1,3 @@
+import { defineMeta } from "blume";
+
+export default defineMeta({ title: "Binary Search", order: 5, icon: "search", display: "group" });
@@ -1,12 +1,13 @@
 ---
 title: '1. Two Sum'
 description: You are given an array of integers nums and an integer target, return indices of the two numbers such that they add up to target
+icon: dot
 sidebar:
  label: 'Two Sum'
  badge: 'Easy'
 ---

-<Badge variant="accent">Hash Table</Badge>
+<Badge variant="accent">Array</Badge>

 ::::warning
 Can you come up with an algorithm that is less than O(n^2) time complexity?
@@ -1,6 +1,7 @@
 ---
 title: '1832. Check if the Sentence Is Pangram'
 description: A pangram is a sentence where every letter of the English alphabet appears at least once
+icon: dot
 sidebar:
  label: 'Check if the Sentence Is Pangram'
  badge: 'Easy'
@@ -1,12 +1,13 @@
 ---
 title: '217. Contains Duplicate'
 description: Given an integer array nums, return true if any value appears at least twice in the array, and return false if every element is distinct
+icon: dot
 sidebar:
  label: 'Contains Duplicate'
  badge: 'Easy'
 ---

-<Badge variant="accent">Hash Table</Badge>
+<Badge variant="accent">Array</Badge>

 ### Example 1:
 - Input: `nums = [1,2,3,1]`
@@ -42,32 +43,12 @@ flowchart TD

 ## Solution

-<CodeGroup>
-
-```py Python
+```py
 class Solution:
    def containsDuplicate(self, nums: List[int]) -> bool:
+        return len(set(nums)) != len(nums)
 ```

-```js JavaScript
-/**
- * @param {number[]} nums
- * @return {boolean}
- */
-var containsDuplicate = function(nums) {
-  const freq = {};
-  for (let n of nums) {
-    freq[n] = (freq[n] || 0) + 1;
-    if (freq[n] >= 2) {
-      return true;
-    }
-  } 
-  return false;
-};
-```
-
-</CodeGroup>
-
 ## Explanation

 <YouTube id="3OamzN90kPg" title="Contains Duplicate - Leetcode 217 - Python" />
@@ -1,6 +1,7 @@
 ---
 title: '242. Valid Anagram'
 description: Given two strings s and t, return true if t is an anagram of s, and false otherwise
+icon: dot
 sidebar:
  label: 'Valid Anagram'
  badge: 'Easy'
@@ -1,12 +1,13 @@
 ---
 title: '347. Top K Frequent Elements'
 description: Given an integer array nums and an integer k, return the k most frequent elements. You may return the answer in any order
+icon: dot
 sidebar:
  label: 'Top K Frequent Elements'
  badge: 'Medium'
 ---

-<Badge variant="accent">Hash Table</Badge>
+<Badge variant="accent">Array</Badge>

 ::::warning
 Your algorithm's time complexity must be better than O(n log n), where n is the array's size.
@@ -1,6 +1,7 @@
 ---
 title: '387. First Unique Character in a String'
 description: Given a string s, find the first non-repeating character in it and return its index. If it does not exist, return -1
+icon: dot
 sidebar:
  label: 'First Unique Character in a String'
  badge: 'Easy'
@@ -1,6 +1,7 @@
 ---
 title: '451. Sort Characters By Frequency'
 description: Given a string s, sort it in decreasing order based on the frequency of the characters. The frequency of a character is the number of times it appears in the string
+icon: dot
 sidebar:
  label: 'Sort Characters By Frequency'
  badge: 'Medium'
@@ -1,12 +1,13 @@
 ---
 title: '49. Group Anagrams'
 description: Given an array of strings strs, group the anagrams together. You can return the answer in any order
+icon: dot
 sidebar:
  label: 'Group Anagrams'
  badge: 'Medium'
 ---

-<Badge variant="accent">Hash Table</Badge>
+<Badge variant="accent">Array</Badge>

 ### Example 1:
 - Input: `strs = ["eat","tea","tan","ate","nat","bat"]`
@@ -0,0 +1,5 @@
+---
+title: Hash Table
+description: 'A key maps to a value in constant time, which removes the need for a full scan.'
+hidden: true
+---
@@ -0,0 +1,3 @@
+import { defineMeta } from "blume";
+
+export default defineMeta({ title: "Hash Table", order: 1, icon: "hash", display: "group" });
@@ -0,0 +1,3 @@
+import { defineMeta } from "blume";
+
+export default defineMeta({ title: "Phase 1: Linear Structures", order: 1 });
@@ -1,12 +1,13 @@
 ---
 title: '739. Daily Temperatures'
 description: Given an array of integers temperatures represents the daily temperatures, return an array answer such that answer[i] is the number of days you have to wait after the i^th day to get a warmer temperature. If there is no future day for which this is possible, keep answer[i] == 0 instead
+icon: dot
 sidebar:
  label: 'Daily Temperatures'
  badge: 'Medium'
 ---

-<Badge variant="accent">Monotonic Stack</Badge>
+<Badge variant="accent">Array</Badge>

 ### Example 1:
 - Input: `temperatures = [73,74,75,71,69,72,76,73]`
@@ -1,12 +1,13 @@
 ---
 title: '853. Car Fleet'
 description: There are n cars at given miles away from the starting mile 0, traveling to reach the mile target
+icon: dot
 sidebar:
  label: 'Car Fleet'
  badge: 'Medium'
 ---

-<Badge variant="accent">Monotonic Stack</Badge>
+<Badge variant="accent">Array</Badge>

 ### Example 1:
 - Input: `target = 12, position = [10,8,0,5,3], speed = [2,4,1,1,3]`
@@ -0,0 +1,5 @@
+---
+title: Monotonic Stack
+description: 'A stack keeps its items in sorted order, which finds the next larger or smaller item in one pass.'
+hidden: true
+---
@@ -0,0 +1,3 @@
+import { defineMeta } from "blume";
+
+export default defineMeta({ title: "Monotonic Stack", order: 7, icon: "trending-up", display: "group" });
@@ -1,12 +1,13 @@
 ---
 title: '238. Product of Array Except Self'
 description: Given an integer array nums, return an array answer such that answer[i] is equal to the product of all the elements of nums except nums[i]
+icon: dot
 sidebar:
  label: 'Product of Array Except Self'
  badge: 'Medium'
 ---

-<Badge variant="accent">Prefix Sum</Badge>
+<Badge variant="accent">Array</Badge>

 ::::warning
 You must write an algorithm that runs in O(n) time and without using the division operation.
@@ -1,12 +1,13 @@
 ---
 title: '303. Range Sum Query - Immutable'
 description: Given an integer array nums, handle multiple queries of the following type
+icon: dot
 sidebar:
  label: 'Range Sum Query - Immutable'
  badge: 'Easy'
 ---

-<Badge variant="accent">Prefix Sum</Badge>
+<Badge variant="accent">Array</Badge>

 ### Example 1:
 - Input: `["NumArray", "sumRange", "sumRange", "sumRange"]` `[[[-2, 0, 3, -5, 2, -1]], [0, 2], [2, 5], [0, 5]]`
@@ -1,12 +1,13 @@
 ---
 title: '560. Subarray Sum Equals K'
 description: Given an array of integers nums and an integer k, return the total number of subarrays whose sum equals to k
+icon: dot
 sidebar:
  label: 'Subarray Sum Equals K'
  badge: 'Medium'
 ---

-<Badge variant="accent">Prefix Sum</Badge>
+<Badge variant="accent">Array</Badge>

 ### Example 1:
 - Input: `nums = [1,1,1], k = 2`
@@ -0,0 +1,5 @@
+---
+title: Prefix Sum
+description: 'A running total is stored at each index, so the sum of any range is one subtraction.'
+hidden: true
+---
@@ -0,0 +1,3 @@
+import { defineMeta } from "blume";
+
+export default defineMeta({ title: "Prefix Sum", order: 2, icon: "sigma", display: "group" });
@@ -1,12 +1,13 @@
 ---
 title: '121. Best Time to Buy and Sell Stock'
 description: You are given an array prices where prices[i] is the price of a given stock on the i^th day
+icon: dot
 sidebar:
  label: 'Best Time to Buy and Sell Stock'
  badge: 'Easy'
 ---

-<Badge variant="accent">Sliding Window</Badge>
+<Badge variant="accent">Array</Badge>

 ### Example 1:
 - Input: `prices = [7,1,5,3,6,4]`
@@ -1,12 +1,13 @@
 ---
 title: '3. Longest Substring Without Repeating Characters'
 description: Given a string s, find the length of the longest substring without duplicate characters
+icon: dot
 sidebar:
  label: 'Longest Substring Without Repeating Characters'
  badge: 'Medium'
 ---

-<Badge variant="accent">Sliding Window</Badge>
+<Badge variant="accent">Hash Table</Badge>

 ### Example 1:
 - Input: `s = "abcabcbb"`
@@ -1,12 +1,13 @@
 ---
 title: '424. Longest Repeating Character Replacement'
 description: You are given a string s and an integer k. You can choose any character of the string and change it to any other uppercase English character. You can perform this operation at most k times
+icon: dot
 sidebar:
  label: 'Longest Repeating Character Replacement'
  badge: 'Medium'
 ---

-<Badge variant="accent">Sliding Window</Badge>
+<Badge variant="accent">Hash Table</Badge>

 ### Example 1:
 - Input: `s = "ABAB", k = 2`
@@ -1,12 +1,13 @@
 ---
 title: '567. Permutation in String'
 description: Given two strings s1 and s2, return true if s2 contains a permutation of s1, or false otherwise
+icon: dot
 sidebar:
  label: 'Permutation in String'
  badge: 'Medium'
 ---

-<Badge variant="accent">Sliding Window</Badge>
+<Badge variant="accent">Hash Table</Badge>

 ### Example 1:
 - Input: `s1 = "ab", s2 = "eidbaooo"`
@@ -1,6 +1,7 @@
 ---
 title: Sliding Window
 description: 'Sliding window is a specialized two-pointer family where the pointers define a contiguous interval and you maintain information about that entire interval as it moves.'
+hidden: true
 ---

 :::warning
@@ -54,7 +55,7 @@ Without an invariant, shrinking the window becomes arbitrary. You no longer know
 The window size never changes. The main task is updating the window state efficiently when one element enters and one leaves.
 :::

-![Sliding Window Fixed Size](../../public/sliding-window-fixed-size.svg)
+![Sliding Window Fixed Size](../../../public/sliding-window-fixed-size.svg)

 ### 2. Flexible-size windows with a validity rule

@@ -62,7 +63,7 @@ The window size never changes. The main task is updating the window state effici
 Here the window expands until invalid, then shrinks until valid again.
 :::

-![Flexible-size windows](../../public/sliding-window-flexible.svg)
+![Flexible-size windows](../../../public/sliding-window-flexible.svg)


 ### 3. Frequency-tracking windows
@@ -71,7 +72,7 @@ Here the window expands until invalid, then shrinks until valid again.
 The difficulty comes from keeping the bookkeeping correct as characters enter and leave.
 :::

-![Frequency-tracking windows](../../public/sliding-window-frequency.svg)
+![Frequency-tracking windows](../../../public/sliding-window-frequency.svg)

 ## How Sliding Window Differs From Other Two-Pointer Families

@@ -0,0 +1,3 @@
+import { defineMeta } from "blume";
+
+export default defineMeta({ title: "Sliding Window", order: 4, icon: "app-window", display: "group" });
@@ -1,12 +1,13 @@
 ---
 title: '150. Evaluate Reverse Polish Notation'
 description: You are given an array of strings tokens that represents an arithmetic expression in a Reverse Polish Notation
+icon: dot
 sidebar:
  label: 'Evaluate Reverse Polish Notation'
  badge: 'Medium'
 ---

-<Badge variant="accent">Stack</Badge>
+<Badge variant="accent">Array</Badge>

 ### Example 1:
 - Input: `tokens = ["2","1","+","3","*"]`
@@ -1,6 +1,7 @@
 ---
 title: '155. Min Stack'
 description: Design a stack that supports push, pop, top, and retrieving the minimum element in constant time
+icon: dot
 sidebar:
  label: 'Min Stack'
  badge: 'Medium'
@@ -1,12 +1,13 @@
 ---
 title: '20. Valid Parentheses'
 description: Given a string s containing just the characters '(', ')', '{', '}', '[' and ']', determine if the input string is valid
+icon: dot
 sidebar:
  label: 'Valid Parentheses'
  badge: 'Easy'
 ---

-<Badge variant="accent">Stack</Badge>
+<Badge variant="accent">String</Badge>

 ### Example 1:
 - Input: `s = "()"`
@@ -0,0 +1,5 @@
+---
+title: Stack
+description: 'Items are added and removed from one end, so the last item added is the first item removed.'
+hidden: true
+---
@@ -0,0 +1,3 @@
+import { defineMeta } from "blume";
+
+export default defineMeta({ title: "Stack", order: 6, icon: "layers", display: "group" });
@@ -1,12 +1,13 @@
 ---
 title: '11. Container With Most Water'
 description: You are given an integer array height of length n. There are n vertical lines drawn such that the two endpoints of the i^th line are (i, 0) and (i, height[i])
+icon: dot
 sidebar:
  label: 'Container With Most Water'
  badge: 'Medium'
 ---

-<Badge variant="accent">Two Pointers</Badge>
+<Badge variant="accent">Array</Badge>

 ::::warning
 Notice that you may not slant the container.
@@ -1,6 +1,7 @@
 ---
 title: '125. Valid Palindrome'
 description: A phrase is a palindrome if, after converting all uppercase letters into lowercase letters and removing all non-alphanumeric characters, it reads the same forward and backward. Alphanumeric characters include letters and numbers
+icon: dot
 sidebar:
  label: 'Valid Palindrome'
  badge: 'Easy'
@@ -1,12 +1,13 @@
 ---
 title: '15. 3Sum'
 description: Given an integer array nums, return all the triplets [nums[i], nums[j], nums[k]] such that i != j, i != k, and j != k, and nums[i] + nums[j] + nums[k] == 0
+icon: dot
 sidebar:
  label: '3Sum'
  badge: 'Medium'
 ---

-<Badge variant="accent">Two Pointers</Badge>
+<Badge variant="accent">Array</Badge>

 ::::warning
 Notice that the solution set must not contain duplicate triplets.
@@ -1,12 +1,13 @@
 ---
 title: '167. Two Sum II - Input Array Is Sorted'
 description: Given a 1-indexed array of integers numbers that is already sorted in non-decreasing order, find two numbers such that they add up to a specific target number. Let these two numbers be numbers[index1] and numbers[index2] where 1 <= index1 < index2 <= numbers.length
+icon: dot
 sidebar:
  label: 'Two Sum II - Input Array Is Sorted'
  badge: 'Medium'
 ---

-<Badge variant="accent">Two Pointers</Badge>
+<Badge variant="accent">Array</Badge>

 ::::warning
 Your solution must use only constant extra space.
@@ -1,6 +1,7 @@
 ---
 title: '344. Reverse String'
 description: Write a function that reverses a string. The input string is given as an array of characters s
+icon: dot
 sidebar:
  label: 'Reverse String'
  badge: 'Easy'
@@ -1,6 +1,7 @@
 ---
 title: '392. Is Subsequence'
 description: Given two strings s and t, return true if s is a subsequence of t, or false otherwise
+icon: dot
 sidebar:
  label: 'Is Subsequence'
  badge: 'Easy'
@@ -1,12 +1,13 @@
 ---
 title: '977. Squares of a Sorted Array'
 description: Given an integer array nums sorted in non-decreasing order, return an array of the squares of each number sorted in non-decreasing order
+icon: dot
 sidebar:
  label: 'Squares of a Sorted Array'
  badge: 'Easy'
 ---

-<Badge variant="accent">Two Pointers</Badge>
+<Badge variant="accent">Array</Badge>

 ::::warning
 Squaring each element and sorting the new array is very trivial, could you find an O(n) solution using a different approach?
@@ -0,0 +1,5 @@
+---
+title: Two Pointers
+description: 'Two indices move through the data from different positions and meet at the answer.'
+hidden: true
+---
@@ -0,0 +1,3 @@
+import { defineMeta } from "blume";
+
+export default defineMeta({ title: "Two Pointers", order: 3, icon: "arrow-left-right", display: "group" });
@@ -1,12 +1,13 @@
 ---
 title: '146. LRU Cache'
 description: Design a data structure that follows the constraints of a Least Recently Used (LRU) cache
+icon: dot
 sidebar:
  label: 'LRU Cache'
  badge: 'Medium'
 ---

-<Badge variant="accent">Hybrid Structures</Badge>
+<Badge variant="accent">Hash Table</Badge>

 ### Example 1:
 - Input: ``
@@ -1,12 +1,13 @@
 ---
 title: '2. Add Two Numbers'
 description: You are given two non-empty linked lists representing two non-negative integers. The digits are stored in reverse order, and each of their nodes contains a single digit. Add the two numbers and return the sum as a linked list
+icon: dot
 sidebar:
  label: 'Add Two Numbers'
  badge: 'Medium'
 ---

-<Badge variant="accent">Hybrid Structures</Badge>
+<Badge variant="accent">Linked List</Badge>

 ### Example 1:
 - Input: `l1 = [2,4,3], l2 = [5,6,4]`
@@ -27,6 +28,20 @@ sidebar:
 - `0 <= Node.val <= 9`
 - It is guaranteed that the list represents a number that does not have leading zeros.

+## Approach
+
+```mermaid
+flowchart TD
+  S(["addTwoNumbers(l1, l2)"]) --> I["dummy = ListNode(), carry = 0, curr = dummy — dummy.next will be the head"]
+  I --> W{"l1 or l2 or carry?"}
+  W -- no --> E(["return dummy.next"])
+  W -- yes --> A["s = (l1.val if l1 else 0) + (l2.val if l2 else 0) + carry — a missing digit counts as 0"]
+  A --> D["carry, val = divmod(s, 10)"]
+  D --> N["curr.next = ListNode(val), curr = curr.next"]
+  N --> V["l1 = l1.next if l1 else None, l2 = l2.next if l2 else None"]
+  V --> W
+```
+
 ## Solution

 ```py
@@ -55,3 +70,7 @@ class Solution:

        return dummy.next
 ```
+
+## Explanation
+
+<YouTube id="wgFPrzTjm7s" title="Add Two Numbers - Leetcode 2 - Python" />
@@ -0,0 +1,5 @@
+---
+title: Hybrid Structures
+description: 'Two structures are combined, such as a hash map and a linked list, to get the best properties of each.'
+hidden: true
+---
@@ -0,0 +1,3 @@
+import { defineMeta } from "blume";
+
+export default defineMeta({ title: "Hybrid Structures", order: 9, icon: "combine", display: "group" });
@@ -1,12 +1,13 @@
 ---
 title: '141. Linked List Cycle'
 description: Given head, the head of a linked list, determine if the linked list has a cycle in it
+icon: dot
 sidebar:
  label: 'Linked List Cycle'
  badge: 'Easy'
 ---

-<Badge variant="accent">Linked List</Badge>
+<Badge variant="accent">Hash Table</Badge>

 ::::warning
 Can you solve it using O(1) (i.e. constant) memory?
@@ -1,6 +1,7 @@
 ---
 title: '206. Reverse Linked List'
 description: Given the head of a singly linked list, reverse the list, and return the reversed list
+icon: dot
 sidebar:
  label: 'Reverse Linked List'
  badge: 'Easy'
@@ -1,6 +1,7 @@
 ---
 title: '21. Merge Two Sorted Lists'
 description: You are given the heads of two sorted linked lists list1 and list2
+icon: dot
 sidebar:
  label: 'Merge Two Sorted Lists'
  badge: 'Easy'
@@ -0,0 +1,5 @@
+---
+title: Linked List
+description: 'Each node points to the next node, and the pointers give the path through the data.'
+hidden: true
+---
@@ -0,0 +1,3 @@
+import { defineMeta } from "blume";
+
+export default defineMeta({ title: "Linked List", order: 8, icon: "link", display: "group" });
@@ -1,12 +1,13 @@
 ---
 title: '48. Rotate Image'
 description: You are given an n x n 2D matrix representing an image, rotate the image by 90 degrees (clockwise)
+icon: dot
 sidebar:
  label: 'Rotate Image'
  badge: 'Medium'
 ---

-<Badge variant="accent">Matrix Index Math</Badge>
+<Badge variant="accent">Array</Badge>

 ### Example 1:
 - Input: `matrix = [[1,2,3],[4,5,6],[7,8,9]]`
@@ -22,6 +23,22 @@ sidebar:
 - `1 <= n <= 20`
 - `-1000 <= matrix[i][j] <= 1000`

+## Approach
+
+```mermaid
+flowchart TD
+  S(["rotate(matrix)"]) --> N["n = len(matrix)"]
+  N --> F{"more r in range(n)?"}
+  F -- yes --> G{"more c in range(r + 1, n)?"}
+  G -- yes --> T["swap matrix[r][c] and matrix[c][r] — transpose across the diagonal"]
+  T --> G
+  G -- no --> F
+  F -- no --> H{"more r in range(n)?"}
+  H -- yes --> R["matrix[r].reverse() — transpose then mirror is a clockwise turn"]
+  R --> H
+  H -- no --> E(["return None — rotated in place"])
+```
+
 ## Solution

 ```py
@@ -35,3 +52,7 @@ class Solution:
        for r in range(n):
            matrix[r].reverse()
 ```
+
+## Explanation
+
+<YouTube id="fMSJSS7eO1w" title="Rotate Image - Matrix - Leetcode 48" />
@@ -1,12 +1,13 @@
 ---
 title: '54. Spiral Matrix'
 description: Given an m x n matrix, return all elements of the matrix in spiral order
+icon: dot
 sidebar:
  label: 'Spiral Matrix'
  badge: 'Medium'
 ---

-<Badge variant="accent">Matrix Index Math</Badge>
+<Badge variant="accent">Array</Badge>

 ### Example 1:
 - Input: `matrix = [[1,2,3],[4,5,6],[7,8,9]]`
@@ -23,6 +24,24 @@ sidebar:
 - `1 <= m, n <= 10`
 - `-100 <= matrix[i][j] <= 100`

+## Approach
+
+```mermaid
+flowchart TD
+  S(["spiralOrder(matrix)"]) --> I["res = [], top = 0, bottom = len(matrix) - 1, left = 0, right = len(matrix[0]) - 1"]
+  I --> W{"top <= bottom and left <= right?"}
+  W -- no --> E(["return res"])
+  W -- yes --> T["append matrix[top][c] for c in left..right, then top += 1 — top row, shrink from above"]
+  T --> R["append matrix[r][right] for r in top..bottom, then right -= 1 — right column"]
+  R --> B{"top <= bottom?"}
+  B -- yes --> C["append matrix[bottom][c] for c in right down to left, then bottom -= 1 — bottom row"]
+  C --> L{"left <= right?"}
+  B -- no --> L
+  L -- yes --> D["append matrix[r][left] for r in bottom down to top, then left += 1 — left column"]
+  D --> W
+  L -- no --> W
+```
+
 ## Solution

 ```py
@@ -57,3 +76,7 @@ class Solution:

        return res
 ```
+
+## Explanation
+
+<YouTube id="BJnMZNwUk1M" title="Spiral Matrix - Microsoft Interview Question - Leetcode 54" />
@@ -1,12 +1,13 @@
 ---
 title: '73. Set Matrix Zeroes'
 description: Given an m x n integer matrix matrix, if an element is 0, set its entire row and column to 0's
+icon: dot
 sidebar:
  label: 'Set Matrix Zeroes'
  badge: 'Medium'
 ---

-<Badge variant="accent">Matrix Index Math</Badge>
+<Badge variant="accent">Array</Badge>

 ::::warning
 You must do it in place.
@@ -30,6 +31,35 @@ You must do it in place.
 - A simple improvement uses O(m + n) space, but still not the best solution.
 - Could you devise a constant space solution?

+## Approach
+
+```mermaid
+flowchart TD
+  S(["setZeroes(matrix)"]) --> I["m = len(matrix), n = len(matrix[0])"]
+  I --> Z["row0_zero = any(matrix[0][c] == 0), col0_zero = any(matrix[r][0] == 0) — remember before they become markers"]
+  Z --> F{"more r in range(1, m)?"}
+  F -- yes --> G{"more c in range(1, n)?"}
+  G -- yes --> Q{"matrix[r][c] == 0?"}
+  Q -- yes --> M["matrix[r][0] = 0, matrix[0][c] = 0 — row 0 and column 0 are the marker lists"]
+  M --> G
+  Q -- no --> G
+  G -- no --> F
+  F -- no --> H{"more r in range(1, m)?"}
+  H -- yes --> J{"more c in range(1, n)?"}
+  J -- yes --> K{"matrix[r][0] == 0 or matrix[0][c] == 0?"}
+  K -- yes --> W["matrix[r][c] = 0"]
+  W --> J
+  K -- no --> J
+  J -- no --> H
+  H -- no --> A{"row0_zero?"}
+  A -- yes --> B["matrix[0][c] = 0 for c in range(n)"]
+  B --> C{"col0_zero?"}
+  A -- no --> C
+  C -- yes --> D["matrix[r][0] = 0 for r in range(m)"]
+  D --> E(["return None — zeroed in place"])
+  C -- no --> E
+```
+
 ## Solution

 ```py
@@ -65,3 +95,7 @@ class Solution:
            for r in range(m):
                matrix[r][0] = 0
 ```
+
+## Explanation
+
+<YouTube id="T41rL0L3Pnw" title="Set Matrix Zeroes - In-place - Leetcode 73" />
@@ -0,0 +1,5 @@
+---
+title: Matrix Index Math
+description: 'A two-dimensional grid is read as a flat array through a formula, and the neighbors are found by arithmetic.'
+hidden: true
+---
@@ -0,0 +1,3 @@
+import { defineMeta } from "blume";
+
+export default defineMeta({ title: "Matrix Index Math", order: 10, icon: "grid-3x3", display: "group" });
@@ -0,0 +1,3 @@
+import { defineMeta } from "blume";
+
+export default defineMeta({ title: "Phase 2: Nodal & Grid", order: 2 });
@@ -1,12 +1,13 @@
 ---
 title: '230. Kth Smallest Element in a BST'
 description: Given the root of a binary search tree, and an integer k, return the k^th smallest value (1-indexed) of all the values of the nodes in the tree
+icon: dot
 sidebar:
  label: 'Kth Smallest Element in a BST'
  badge: 'Medium'
 ---

-<Badge variant="accent">Binary Search Tree</Badge>
+<Badge variant="accent">Tree</Badge>

 ::::warning
 If the BST is modified often (i.e., we can do insert and delete operations) and you need to find the kth smallest frequently, how would you optimize?
@@ -26,6 +27,23 @@ If the BST is modified often (i.e., we can do insert and delete operations) and
 - `1 <= k <= n <= 10^4`
 - `0 <= Node.val <= 10^4`

+## Approach
+
+```mermaid
+flowchart TD
+  S(["kthSmallest(root, k)"]) --> I["n = 0, stack = [], cur = root"]
+  I --> W{"cur or stack?"}
+  W -- no --> E(["fall off the loop — unreachable, k <= n"])
+  W -- yes --> L{"cur?"}
+  L -- yes --> P["stack.append(cur), cur = cur.left — push the whole left spine"]
+  P --> L
+  L -- no --> O["cur = stack.pop(), n += 1 — nodes pop in ascending order"]
+  O --> K{"n == k?"}
+  K -- yes --> R(["return cur.val"])
+  K -- no --> N["cur = cur.right"]
+  N --> W
+```
+
 ## Solution

 ```py
@@ -52,3 +70,7 @@ class Solution:
                return cur.val
            cur = cur.right
 ```
+
+## Explanation
+
+<YouTube id="5LUXSvjmGCw" title="Kth Smallest Element in a BST - Leetcode 230 - Python" />
@@ -1,12 +1,13 @@
 ---
 title: '235. Lowest Common Ancestor of a Binary Search Tree'
 description: Given a binary search tree (BST), find the lowest common ancestor (LCA) node of two given nodes in the BST
+icon: dot
 sidebar:
  label: 'Lowest Common Ancestor of a Binary Search Tree'
  badge: 'Medium'
 ---

-<Badge variant="accent">Binary Search Tree</Badge>
+<Badge variant="accent">Tree</Badge>

 ### Example 1:
 - Input: `root = [6,2,8,0,4,7,9,null,null,3,5], p = 2, q = 8`
@@ -30,6 +31,22 @@ sidebar:
 - `p != q`
 - p and q will exist in the BST.

+## Approach
+
+```mermaid
+flowchart TD
+  S(["lowestCommonAncestor(root, p, q)"]) --> I["cur = root"]
+  I --> W{"cur?"}
+  W -- no --> E(["fall off the loop — unreachable, p and q exist in the BST"])
+  W -- yes --> A{"p.val < cur.val and q.val < cur.val?"}
+  A -- yes --> L["cur = cur.left — both sit in the left subtree"]
+  L --> W
+  A -- no --> B{"p.val > cur.val and q.val > cur.val?"}
+  B -- yes --> R["cur = cur.right — both sit in the right subtree"]
+  R --> W
+  B -- no --> C(["return cur — the paths to p and q split here"])
+```
+
 ## Solution

 ```py
@@ -54,3 +71,7 @@ class Solution:
            else:
                return cur
 ```
+
+## Explanation
+
+<YouTube id="gs2LMfuOR9k" title="Lowest Common Ancestor of a Binary Search Tree - Leetcode 235 - Python" />
@@ -1,12 +1,13 @@
 ---
 title: '98. Validate Binary Search Tree'
 description: Given the root of a binary tree, determine if it is a valid binary search tree (BST)
+icon: dot
 sidebar:
  label: 'Validate Binary Search Tree'
  badge: 'Medium'
 ---

-<Badge variant="accent">Binary Search Tree</Badge>
+<Badge variant="accent">Tree</Badge>

 ### Example 1:
 - Input: `root = [2,1,3]`
@@ -22,6 +23,23 @@ sidebar:
 - The number of nodes in the tree is in the range [1, 10^4].
 - `-2^31 <= Node.val <= 2^31 - 1`

+## Approach
+
+```mermaid
+flowchart TD
+  S(["isValidBST(root)"]) --> C["valid(root, float('-inf'), float('inf')) — the root may hold any value"]
+  C --> E(["return that result"])
+  subgraph valid["valid(node, low, high)"]
+    V1{"not node?"}
+    V1 -- yes --> V2(["return True"])
+    V1 -- no --> V3{"not (low < node.val < high)?"}
+    V3 -- yes --> V4(["return False — node breaks a bound set by an ancestor"])
+    V3 -- no --> V5["valid(node.left, low, node.val) and valid(node.right, node.val, high) — each child tightens one bound"]
+    V5 --> V6(["return that result"])
+  end
+  C --> V1
+```
+
 ## Solution

 ```py
@@ -44,3 +62,7 @@ class Solution:

        return valid(root, float("-inf"), float("inf"))
 ```
+
+## Explanation
+
+<YouTube id="s6ATEkipzow" title="Validate Binary Search Tree - Depth First Search - Leetcode 98" />
@@ -0,0 +1,5 @@
+---
+title: Binary Search Tree
+description: 'Each node holds a value that is larger than its left subtree and smaller than its right subtree.'
+hidden: true
+---
@@ -0,0 +1,3 @@
+import { defineMeta } from "blume";
+
+export default defineMeta({ title: "Binary Search Tree", order: 13, icon: "git-fork", display: "group" });
@@ -1,12 +1,13 @@
 ---
 title: '215. Kth Largest Element in an Array'
 description: Given an integer array nums and an integer k, return the k^th largest element in the array
+icon: dot
 sidebar:
  label: 'Kth Largest Element in an Array'
  badge: 'Medium'
 ---

-<Badge variant="accent">Heap / Priority Queue</Badge>
+<Badge variant="accent">Array</Badge>

 ### Example 1:
 - Input: `nums = [3,2,1,5,6,4], k = 2`
@@ -21,6 +22,20 @@ sidebar:
 - `1 <= k <= nums.length <= 10^5`
 - `-10^4 <= nums[i] <= 10^4`

+## Approach
+
+```mermaid
+flowchart TD
+  S(["findKthLargest(nums, k)"]) --> F{"more i in range(len(nums))?"}
+  F -- yes --> N["nums[i] = -nums[i] — negate so the min-heap acts as a max-heap"]
+  N --> F
+  F -- no --> H["heapq.heapify(nums)"]
+  H --> W{"more _ in range(k - 1)?"}
+  W -- yes --> P["heapq.heappop(nums) — discard the k-1 largest"]
+  P --> W
+  W -- no --> E(["return -heapq.heappop(nums)"])
+```
+
 ## Solution

 ```py
@@ -39,3 +54,7 @@ class Solution:

        return -heapq.heappop(nums)
 ```
+
+## Explanation
+
+<YouTube id="XEmy13g1Qxc" title="Kth Largest Element in an Array - Quick Select - Leetcode 215 - Python" />
@@ -1,12 +1,13 @@
 ---
 title: '23. Merge k Sorted Lists'
 description: You are given an array of k linked-lists lists, each linked-list is sorted in ascending order
+icon: dot
 sidebar:
  label: 'Merge k Sorted Lists'
  badge: 'Hard'
 ---

-<Badge variant="accent">Heap / Priority Queue</Badge>
+<Badge variant="accent">Linked List</Badge>

 ### Example 1:
 - Input: `lists = [[1,4,5],[1,3,4],[2,6]]`
@@ -30,6 +31,27 @@ sidebar:
 - `lists[i]` is sorted in ascending order.
 - The sum of lists[i].length will not exceed 10^4.

+## Approach
+
+```mermaid
+flowchart TD
+  S(["mergeKLists(lists)"]) --> I["heap = []"]
+  I --> F{"more i, node in enumerate(lists)?"}
+  F -- yes --> C{"node?"}
+  C -- yes --> P["heapq.heappush(heap, (node.val, i, node)) — i breaks ties so nodes are never compared"]
+  P --> F
+  C -- no --> F
+  F -- no --> D["D = ListNode(), cur = D"]
+  D --> W{"heap?"}
+  W -- no --> E(["return D.next"])
+  W -- yes --> O["val, i, node = heapq.heappop(heap) — the smallest head across all lists"]
+  O --> L["cur.next = node, cur = node, node = node.next"]
+  L --> N{"node?"}
+  N -- yes --> Q["heapq.heappush(heap, (node.val, i, node))"]
+  Q --> W
+  N -- no --> W
+```
+
 ## Solution

 ```py
@@ -67,3 +89,7 @@ class Solution:
        # Space: O(n)
        return D.next
 ```
+
+## Explanation
+
+<YouTube id="q5a5OiGbT6Q" title="Merge K Sorted Lists - Leetcode 23 - Python" />
@@ -1,12 +1,13 @@
 ---
 title: '973. K Closest Points to Origin'
 description: Given an array of points where points[i] = [xi, yi] represents a point on the X-Y plane and an integer k, return the k closest points to the origin (0, 0)
+icon: dot
 sidebar:
  label: 'K Closest Points to Origin'
  badge: 'Medium'
 ---

-<Badge variant="accent">Heap / Priority Queue</Badge>
+<Badge variant="accent">Array</Badge>

 ### Example 1:
 - Input: `points = [[1,3],[-2,2]], k = 1`
@@ -23,6 +24,23 @@ sidebar:
 - `1 <= k <= points.length <= 10^4`
 - `-10^4 <= xi, yi <= 10^4`

+## Approach
+
+```mermaid
+flowchart TD
+  S(["kClosest(points, k)"]) --> I["minHeap = []"]
+  I --> F{"more x, y in points?"}
+  F -- yes --> D["dist = x**2 + y**2 — squared distance orders the same as sqrt"]
+  D --> A["minHeap.append([dist, x, y])"]
+  A --> F
+  F -- no --> H["heapq.heapify(minHeap), res = [] — O(n) build, smallest dist on top"]
+  H --> W{"k > 0?"}
+  W -- yes --> P["dist, x, y = heapq.heappop(minHeap)"]
+  P --> R["res.append([x, y]), k -= 1"]
+  R --> W
+  W -- no --> E(["return res"])
+```
+
 ## Solution

 ```py
@@ -46,3 +64,7 @@ class Solution:

        return res
 ```
+
+## Explanation
+
+<YouTube id="rI2EBUEMfTk" title="K Closest Points to Origin - Heap / Priority Queue - Leetcode 973 - Python" />
@@ -0,0 +1,5 @@
+---
+title: Heap / Priority Queue
+description: 'The largest or smallest item is always at the top, and it is removed in logarithmic time.'
+hidden: true
+---
@@ -0,0 +1,3 @@
+import { defineMeta } from "blume";
+
+export default defineMeta({ title: "Heap / Priority Queue", order: 14, icon: "triangle", display: "group" });
@@ -0,0 +1,3 @@
+import { defineMeta } from "blume";
+
+export default defineMeta({ title: "Phase 3: Hierarchical", order: 3 });
Author	SHA1	Message	Date
Prad Nukala	6f9e4ecb3b	chore: clean up content phase organization	2026-09-25 14:00:22 -04:00
Prad Nukala	5dc2265ee5	chore(config): add GitHub repository metadata to docs config	2026-09-25 13:36:50 -04:00
Prad Nukala	0ca130156d	chore: clean up content phase organization	2026-09-25 13:30:06 -04:00
Prad Nukala	0d64df58f1	chore: clean up content phase organization	2026-09-25 13:27:06 -04:00
Prad Nukala	88d1767184	docs: replace legacy line‑dot‑left‑horizontal icons with unified dot icon across docs	2026-09-25 13:17:35 -04:00
Prad Nukala	93381f2b69	style(cli): change default page icon to "dot"	2026-09-25 13:17:31 -04:00
Prad Nukala	58ab9db1d7	docs(docs): add page icon frontmatter to problem pages	2026-09-25 13:14:25 -04:00
Prad Nukala	e189bb876e	feat(cli): add page icon frontmatter support	2026-09-25 13:14:22 -04:00
Prad Nukala	314f2bff10	docs(docs): rename progress_1.mdx to keywords.mdx	2026-09-25 13:11:13 -04:00
Prad Nukala	082b67c44d	docs(docs): add Keywords reference page	2026-09-25 13:10:59 -04:00
Prad Nukala	ad7c0a2972	docs(readme): update documentation URL to lc.prd.lk	2026-09-24 16:26:18 -04:00
Prad Nukala	931c2c64be	docs(docs): remove /leetcode prefix from homepage tile URLs	2026-09-24 16:26:15 -04:00
Prad Nukala	0a99557aac	feat(config): update deployment site URL and reformat font variant definitions	2026-09-24 16:26:12 -04:00
Prad Nukala	7875b43981	docs(api): update docs URL and origin endpoints	2026-09-24 16:26:10 -04:00
Prad Nukala	f34a9bb56d	ci(git): sync redirects in pre-commit hook	2026-09-24 16:15:59 -04:00
Prad Nukala	bfeadc0cdf	docs(docs): add generated short URL redirects for LeetCode problem pages	2026-09-24 16:15:55 -04:00
Prad Nukala	ca05c32f6e	feat(config): add custom redirects to Blume config	2026-09-24 16:15:52 -04:00
Prad Nukala	ad129b99fd	feat(cli): generate short numeric redirects for problem pages	2026-09-24 16:15:49 -04:00
Prad Nukala	717feb7f70	build(config): switch to local Fontsource fonts for offline Blume builds	2026-09-24 16:07:17 -04:00
Prad Nukala	bf0ac34a43	chore: clean up content phase organization	2026-09-24 15:57:38 -04:00
Prad Nukala	403a2d7f23	chore: clean up obsolete examples	2026-09-24 15:32:21 -04:00
Prad Nukala	3445c47ad2	docs(docs): change homepage icon to play	2026-09-24 15:17:19 -04:00
Prad Nukala	24a9c52260	docs(docs): add hidden index pages for algorithm topics and progress dashboard component	2026-09-24 15:17:08 -04:00
Prad Nukala	9d535c6826	feat(cli): preserve leading numeric prefixes in Blume group slugs	2026-09-24 15:17:04 -04:00
Prad Nukala	a2d32c5f06	docs: add icons to topic meta definitions	2026-09-24 14:45:12 -04:00
Prad Nukala	c69582552b	feat(cli): add sidebar icons mapping for topics	2026-09-24 14:45:09 -04:00
Prad Nukala	f0f479d6cf	feat: add lc 46, 39, 78	2026-09-24 14:36:19 -04:00
Prad Nukala	ac4dc654fc	feat(work): add solution for LeetCode 323 number of connected components problem	2026-09-23 15:40:39 -04:00
Prad Nukala	273148c08b	feat: add lc 684	2026-09-23 15:22:26 -04:00
Prad Nukala	7960dee431	docs(docs): adjust default zoom to 1.6× for improved label readability	2026-09-22 15:10:30 -04:00
Prad Nukala	0009075138	feat(work): implement Course Schedule II solution	2026-09-22 15:02:31 -04:00
Prad Nukala	ca73d3235c	feat(work): add Course Schedule II solution and clean up Course Schedule I implementation	2026-09-22 14:24:38 -04:00
Prad Nukala	97b62f95c9	docs(docs): detail custom MDX components and islands in docs overview	2026-09-22 14:11:37 -04:00
Prad Nukala	1a440e55ce	docs(docs): add Carousel component and embed in topological‑sort page, replace static spine image with interactive component	2026-09-22 14:11:33 -04:00
Prad Nukala	fd3b1952b1	docs(docs): add Topological Sort documentation page with example code and images	2026-09-22 13:28:23 -04:00
Prad Nukala	82c26e2b6c	docs: update README URLs and fix sync bullet formatting	2026-09-22 13:06:27 -04:00
Prad Nukala	f78ecb2bd4	docs(docs): add default order 0 to Array meta	2026-09-22 13:06:23 -04:00
Prad Nukala	b3d3c5fbc9	refactor(cli): default group order to 0 and clarify meta comment	2026-09-22 13:06:20 -04:00
Prad Nukala	99a9248368	chore(githooks): add pre-commit hook to sync docs and stage generated pages	2026-09-22 13:01:32 -04:00
Prad Nukala	2a925f6a9a	docs(githooks): replace pre-push hook with pre-commit and update documentation	2026-09-22 13:01:28 -04:00
Prad Nukala	7d57faa958	feat(work): add solution stub for Course Schedule (207)	2026-09-22 13:01:22 -04:00
Prad Nukala	26085cc027	docs(docs): add Word Ladder and Rotting Oranges documentation pages	2026-09-22 12:59:21 -04:00
Prad Nukala	216152d946	init: hooks	2026-09-22 12:59:13 -04:00
Prad Nukala	d0ace85bbc	docs(agents): add .githooks entry describing git hook enforcement and pre‑push sync check	2026-09-22 12:59:00 -04:00
Prad Nukala	5b99a6a589	feat(work): add Word Ladder and Rotting Oranges solutions	2026-09-22 12:58:56 -04:00
Prad Nukala	8446623607	chore(config): add prepare script for git hooks	2026-09-22 12:58:54 -04:00
Prad Nukala	3a1df27d30	docs(readme): replace static chart images with live SRS API badges and update description.	2026-09-22 11:58:58 -04:00
Prad Nukala	ae1d1ceede	feat(work): implement Pacific Atlantic Water Flow solution	2026-09-21 18:05:35 -04:00
Prad Nukala	f5f31ba1c3	docs(docs): remove array order meta and add solution implementation for Pacific Atlantic Water Flow	2026-09-21 18:05:30 -04:00
Prad Nukala	d1c99c307e	docs(docs): add order to array meta	2026-09-21 17:41:19 -04:00
Prad Nukala	021b1fd06e	docs(agents): update docs for topic folder naming and sync behavior	2026-09-21 17:37:31 -04:00
Prad Nukala	edd37ed0e6	docs: rename category folders to remove parentheses and update heap meta title	2026-09-21 17:37:26 -04:00
Prad Nukala	c2a8a31ede	chore(config): remove Reference tab from navigation	2026-09-21 17:37:22 -04:00
Prad Nukala	891a1f0c5a	refactor(cli): clarify topic folder handling and remove unused groupFolder helper	2026-09-21 17:37:19 -04:00
Prad Nukala	2e5d991d47	chore(config): bump multiple package versions in bun.lock	2026-09-21 17:16:31 -04:00
Prad Nukala	73d63c2565	docs(docs): clarify apps/docs layout and bump blume to ^1.7.3	2026-09-21 17:16:18 -04:00
Prad Nukala	cafc8a4c25	feat(work): add solutions for number-of-islands, pacific-atlantic-water-flow, and clone-graph problems	2026-09-21 17:16:13 -04:00
Prad Nukala	7c5061dcd1	docs(docs): add meta files and reorganize topics with updated badges and directory structure	2026-09-21 17:16:10 -04:00
Prad Nukala	f9e69960c4	feat(cli): add topic grouping with meta.ts generation in sync script	2026-09-21 17:15:56 -04:00
Prad Nukala	0c12df3eba	docs(readme): update LeetCode profile link to progress page	2026-09-21 15:55:59 -04:00
Prad Nukala	bfb53c14e8	docs(readme): wrap LeetCode badge in profile link	2026-09-21 15:54:29 -04:00
Prad Nukala	2c30b3d017	feat(problem-approach): add skill for generating approach diagrams	2026-09-21 15:53:32 -04:00
Prad Nukala	1084410832	docs(readme): add videos command and update schedule dates and badges	2026-09-21 15:53:28 -04:00
Prad Nukala	76c9c77307	feat(work): add Number of Islands solution	2026-09-21 15:53:24 -04:00
Prad Nukala	23eb71d387	docs(docs): add Approach diagrams and Explanation sections to LeetCode problem documentation	2026-09-21 15:53:21 -04:00
Prad Nukala	b8c779903f	feat(config): add videos script to package.json	2026-09-21 15:53:18 -04:00
Prad Nukala	5059eb72cb	feat(cli): add videos script for NeetCode embed lookup and CLI command	2026-09-21 15:53:15 -04:00
Prad Nukala	204e178fa4	feat(work): add solution for LeetCode 211 design add and search words data structure	2026-09-21 12:50:53 -04:00
Prad Nukala	61dc19fb29	docs(docs): add MDX page for LeetCode 211 design problem	2026-09-21 12:50:51 -04:00
Prad Nukala	5c644ca050	docs(schedule): clarify schedule JSON format, gate problem handling, and update week dates in README	2026-09-21 12:14:37 -04:00
Prad Nukala	fb2d4c88c2	feat(work): add solution for LeetCode 208 Implement Trie (Prefix Tree)	2026-09-21 12:14:33 -04:00
Prad Nukala	1fa0731ea7	docs(docs): add MDX page for LeetCode 208 Implement Trie (Prefix Tree)	2026-09-21 12:14:29 -04:00
Prad Nukala	86107d5ef7	feat(api): add structured schedule plan with topics, problems, gate, sweep and reserved problem handling	2026-09-21 12:14:25 -04:00