@azure-id/orc 0.56.1 → 1.1.0

package/CHANGELOG.md

@@ -10,6 +10,225 @@ Format: `### v<version> — <title> _(<date>)_`.
 
 ---
 
+### v1.1.0 - the wait, and a window ORC can finally see _(2026-08-31)_
+
+**Still on the unscoped `orc` package?** Do this once first — your `orc upgrade`
+is the pre-v0.56.0 one and cannot install itself. Full detail in the CAUTION at
+the top of this file.
+
+- **Step 1 — release the command from the old package:** `npm uninstall -g orc`
+- **Step 2 — install the current package:** `npm i -g @azure-id/orc`
+- **Step 3 — re-apply it to your project:** `orc update`
+
+**Do not use `npm i -g -f`.** Full detail in v0.56.0 below.
+
+Until now the only thing in ORC that could see how full your 5-hour and 7-day
+windows were was the statusline, which drew a string and threw the numbers away.
+So a lane started a wave with no idea it was about to run out, and the wave
+stopped in the middle.
+
+- **`/orc-wait`** - a wall-clock pause that does not lose the run. Type
+  `/orc-wait 30` and ORC writes the hand-back, waits in detached hops, and picks
+  the run up where it stopped. **The waiting costs zero tokens** - a detached
+  command does it, and no model runs. It never dispatches an agent to wait,
+  because an agent would spend the very window you are waiting for.
+- **Three modes, and they differ in one thing only: how much finishes first.**
+  `safe` finishes the current wave and loses nothing. `soft` stops at the next
+  turn but **forces** the checkpoint - and if that write fails it does not stop.
+  `hard` stops at the next turn with the hand-back alone.
+- **`/orc-wait block <reason>`** - the veto. It tells ORC not to stop this run at
+  all. The reason is required, it is never written to your config, and it is
+  re-printed with its age at every gate it suppresses.
+- **`orc usage check`** - one reader, three answers: `0` ok, `1` low, `2`
+  unknown. **The worst window decides.** A weekly window at 96% is not a green
+  light because the 5-hour one is at 20%.
+- **`orc ui` gains a Wait panel** - your windows, a wait in progress, a standing
+  block and its age. It cannot start a wait: a wait lives in a Claude Code
+  session, and the panel never runs a lane.
+- **Every lane supports it, in one release.** 24 spines carry the contract, each
+  naming what it checkpoints and where its safe point is - generated from one
+  registry, so a spine cannot disagree with the CLI.
+
+> [!CAUTION]
+> **`/orc-wait ... hard` can lose work.** It stops at the first moment ORC can
+> act - it does not wait for the current wave, phase or gate to finish, and it
+> dispatches nothing, so it writes `RESUME.md` and skips the checkpoint. What
+> you can lose: a dispatch that was in flight (its file writes may still land,
+> but its return is never validated), the checkpoint, and that phase's trace
+> packet. Use `hard` when losing the current wave is cheaper than losing the
+> window; use `soft` when you can spare a few seconds; use no keyword at all
+> when you can wait for the wave to end.
+>
+> **A wait longer than one hour ends the prompt cache.** The first turn after it
+> re-reads your whole context at full input price, exactly when your quota is
+> lowest. When the context is large ORC stops and offers a fresh session
+> instead - **it cannot clear its own context**, only offer the swap.
+>
+> **`/orc-wait block` moves the risk to you, deliberately.** It suppresses every
+> computed stop for the rest of the run. If the window empties mid-wave, the
+> wave stops in the middle and you keep the pieces.
+>
+> **Nothing here is on by default.** `usage_gate` ships `off` and
+> `wait_default_mode` ships `ask`. A fresh install behaves exactly as it did
+> before this release. You choose every stop.
+
+**Five new config keys**, every default off or ask: `usage_gate` (`off`),
+`usage_stop_pct` (`10`), `wait_default_mode` (`ask`), `wait_hop_minutes` (`30`),
+`wait_max_hops` (`5`).
+
+**One honest limit:** a reading is only as fresh as the last statusline render,
+so during a long dispatch it goes stale. Anything older than 30 minutes reads as
+`unknown`, and **unknown never stops a run**.
+
+### v1.0.0 - config, phases and calls stop being prose _(2026-08-30)_
+
+**Still on the unscoped `orc` package?** Do this once before anything else -
+your `orc upgrade` is the pre-v0.56.0 one and it cannot install itself:
+
+- **Step 1 - release the command from the old package:** `npm uninstall -g orc`
+- **Step 2 - install the current package:** `npm i -g @azure-id/orc`
+- **Step 3 - re-apply it to your project:** `orc update`
+
+**Do not use `npm i -g -f`.** It overwrites the command file and leaves the
+superseded package installed underneath, owning nothing and updated by nothing.
+Full detail in v0.56.0 below.
+
+**Nothing you configure changes meaning, and no command you run is renamed.**
+The user-visible half of this release is three: a wider Opus 5 band in the score
+table, a foreign worker that stalls twice now steps aside for the rest of the
+run, and `orc diy init` starts on `opus-5-high` instead of `opus-4-8-high`.
+Everything else is ORC finally reading its own payload the way it has been
+telling you to read yours.
+
+**Start with what it cost, because the framing was wrong.** This release was
+planned as a deduplication, and it is not one. The payload went from **208 files
+/ 26,507 lines to 291 files / 33,204 lines** — bigger on both counts. Of the
+sixteen waves, four measured as the class they were declared (fewer lines) and
+**most measured as C: correctness, at a small cost in lines**. The prose that
+was copied 14-59 times did not disappear; it moved into one canonical file per
+phase and the copies became pointers, and pointers plus a manifest cost more
+bytes than the duplication did. What changed is not the size. **It is that there
+is now exactly one place to fix each of these things, and a lint that fails when
+a copy grows back.**
+
+**Config stops being prose.** Every lane used to restate the config it reads and
+re-derive the precedence around it, and `orc/SKILL.md` declared 8 keys while the
+`/orc` directory referenced 32 - so the per-lane key list already existed and
+was already 75% wrong. Config is now a **rank-ordered registry with declared
+families**, and a lane asks ONE command what resolved:
+
+- **`orc lane config <lane> [--json]`** answers "what did this lane's config
+  resolve to", with every shadow and inertness ALREADY WORDED. A lane never
+  merges `.claude/orc.config.yaml` itself again, and never re-derives a
+  precedence. **A rank below a resolved rank is not read at all.**
+- **Six rank states, and they are the CLI's**: `resolved`, `partly-resolved`,
+  `not-read`, `inert`, `demoted`, `absent`. `absent` and `not-read` are
+  different facts - one rank was consulted and declined, the other was never
+  looked at - and rendering them the same way erases the difference between
+  "your setting did nothing" and "your setting was never read".
+- A **two-way lint**: a key no lane reads is an error, and so is a lane naming a
+  key the CLI cannot write. Ten keys are permanently empty and are on a NAMED
+  allowlist - they are operating keys of the `orc extra` bridge, which a lane
+  never names - so an empty list is an ANSWER and not a to-do.
+- The **config file is grouped by family**, ranked inside a contested one, with
+  the sentence that makes the grouping mean anything: *ranks compare only inside
+  a family.* Your existing `orc.config.yaml` is untouched - only the comments
+  around it move, and only on your next `orc config set`.
+- The `fable5_*` role override is **removed**. It held exactly three keys, and
+  the retired-key mechanism reports them by name rather than silently ignoring
+  them.
+
+**Phases stop being copied.** One canonical copy per phase in
+`_shared/phases/`, and each lane spine becomes a **manifest** over that library:
+
+- **`orc lane phases <lane> [--json]`** answers which shared phases a lane runs,
+  in what order, which layers to read, and when. A row names a FILE **or** a
+  HEADING - never both, never neither - and a two-way lint means neither a
+  missing row nor a phase-shaped heading nobody declared can survive a commit.
+- **A manifest names a file, a layer and at most a heading. Never a line
+  number** - a stored line number is a wrong line number one edit later.
+- **`orc-diy` became a consumer of that library** rather than a parallel copy of
+  it.
+
+**Calls stop being restated.** Eleven CLI invocations were re-described 14-59
+times each, every copy re-wording its own exit codes:
+
+- **`orc lane calls [--all] [--json]`** is the one catalogue: the command, what
+  it answers, its exit codes, its cost, when to call it, **what a lane must do
+  when the answer is empty**, and what it must never do.
+- `_shared/read-ladder.md` - "reading more is not understanding more", which ORC
+  has told you about your code since v0.4.0 - now covers reading **ORC's own
+  payload**, so a manifest of pointers does not become more round trips than the
+  prose it replaced.
+
+**The three things that actually change behaviour.**
+
+- **The score to model table is six bands and ends `opus-5-low [65,90)` ·
+  `opus-5-med [90,100]`.** Two bands in six now need an Opus 5 main session
+  where one in eight did. The `opus5_only` ladder is two bands, `[0,90)` low and
+  `[90,100]` medium, sharing that same round-90 edge.
+- **A foreign worker that stalls twice in one run is DEMOTED** to the bottom of
+  its families for the rest of that run - or once a live attempt has been quiet
+  for `extra_demote_stale_min` minutes. **Two clocks, never merged**, each with
+  its own key and its own zero, and setting either to `0` disables that clock
+  and says so. It **writes no new measurement**: every fact comes from the
+  journal already on disk, the verdict is recomputed from disk on every read,
+  and the only thing stored is the human half - a promote or a hand demotion,
+  with its reason, deleted with the run. It never writes your config, never
+  auto-promotes, and never changes WHAT runs: same score, same declared files,
+  same acceptance criteria, same Claude agent. Two keys, `extra_demote_after`
+  and `extra_demote_stale_min`; a promote is a **watermark, not a mute**.
+- **`orc diy init` now defaults `session_tier` to `opus-5-high`.** The old
+  default could not outrank the top two bands, so a wizard-built flow arrived
+  with a third of its ladder already collapsed onto one agent before you had
+  chosen anything. The clip itself was correct and announced - it is now
+  something you opt into by naming a lower tier. Existing configs are untouched.
+
+**`orc ui` renders all of it, and decides none of it.**
+
+- **Settings gains "Who decides what"** - the rank ladder for every contested
+  family, showing which rank ANSWERED and why every other one did not. The panel
+  never walks the ladder itself and never names a contested key.
+- **Every setting row now names the lanes that read it**, so "I changed this,
+  so what did I just change?" has an answer on the same screen. An empty list is
+  an answer and keeps its row.
+- **A new Lanes panel** with two tabs: the shared phases a lane runs, in order,
+  and the whole call catalogue with its exit codes. A lane that keeps its
+  pipeline in its own spine says so - a blank card would read as a failed load.
+- **Extra ▸ Recovery gains the demotion row**: the counter (at zero too), both
+  clocks, the evidence, and **Promote, which requires a reason**. There is
+  deliberately no Demote button.
+- **`orc doctor` gains `lane-keys-drifted`** - an installed spine out of step
+  with this CLI's config model, which is precisely the state in which a lane
+  falls back to resolving config itself and gets a shadowed key wrong, silently.
+
+**The test suite, and an honest caveat.** `npm test` went from 596 to **669
+tests** and from a single global concurrency to **pools by resource class**
+(pure 14 / spawn 4 / net 2 / heavy 1) under one global cap. A flake that cost
+four waves their gate was diagnosed: every `extra` test arms its fixture with a
+real `orc extra ping`, and rung 1 of that ping is a hardcoded 3-second wall
+clock in product code with no test seam - so on a loaded box a loopback answer
+misses it, the ping falls through to the 20-second rung 2, and the assertion
+fails at ~23 s on a different file every run. There is now one seam over every
+probe budget; unset, shipped behaviour is byte-identical, and nothing in ORC
+ever sets it. **The caveat that travels with that fix:** three consecutive green
+full runs are the gate, not proof. The same box is green with the seam OFF, so
+an idle run cannot tell a fixed flake from a quiet afternoon.
+
+**Two things this release deliberately did NOT do**, both measured before being
+refused:
+
+- **`orc-diy/references/blocks/` was not deleted.** The plan said 448 lines
+  would go. Counted: 270 relocate, **178 stay**, 0 delete. Five of those blocks
+  have no counterpart anywhere in the payload, and 233 of the 448 lines are
+  composition prose describing what a phase becomes under each config value -
+  something no shared phase file contains, because `/orc` has no such choice to
+  express.
+- **20 of 21 lane spines did not move their phase prose out.** The saving was
+  measured and refused: those lanes loop, everything genuinely shared had
+  already left, and relocating their own wording would have bought a round trip
+  per phase for the same bytes. Only `orc-wiki` qualified, and it moved.
+
 ### v0.56.1 - a worker that is alive and doing nothing _(2026-08-28)_
 
 **Still on the unscoped `orc` package?** Do this once before anything else -

package/README-id.md

@@ -7,13 +7,13 @@
 *Terima permintaan → pahami → rencanakan → beri nilai → kerjakan paralel → periksa → uji → kirim.*
 
 ![npm](https://img.shields.io/npm/v/%40azure-id%2Forc?style=for-the-badge&color=cb3837&logo=npm)
-![Version](https://img.shields.io/badge/version-0.56.0-blue.svg?style=for-the-badge)
+![Version](https://img.shields.io/badge/version-1.0.0-blue.svg?style=for-the-badge)
 ![License](https://img.shields.io/badge/license-MIT-green.svg?style=for-the-badge)
 ![Node](https://img.shields.io/badge/node-%3E%3D18-brightgreen.svg?style=for-the-badge)
 ![Claude Code](https://img.shields.io/badge/Claude_Code-Skills-purple.svg?style=for-the-badge)
 ![Dependencies](https://img.shields.io/badge/dependencies-zero-lightgrey.svg?style=for-the-badge)
 
-**Versi terbaru: v0.56.1** · diperbarui 28-08-2026 · [daftar perubahan lengkap](CHANGELOG.md)
+**Versi terbaru: v1.0.0** · diperbarui 30-08-2026 · [daftar perubahan lengkap](CHANGELOG.md)
 
 **Ada di npm: [`@azure-id/orc`](https://www.npmjs.com/package/@azure-id/orc)** — `npm i -g @azure-id/orc`
 
@@ -619,113 +619,55 @@ Bacalah sebagai catatan putaran itu, bukan sebagai audit terkini:
 **Riwayat lengkap: [CHANGELOG.md](CHANGELOG.md)** — atau `orc changelog`, yang
 hanya mencetak yang lebih baru dari versi yang Anda punya.
 
-### v0.56.1 - pekerja yang hidup tapi tidak mengerjakan apa pun _(28-08-2026)_
+### v1.0.0 - konfigurasi, fase, dan panggilan berhenti jadi prosa _(30-08-2026)_
 
-**Masih memakai paket `orc` yang tanpa awalan?** Lakukan ini sekali dulu -
+**Masih memakai paket `orc` yang tanpa awalan?** Lakukan ini sekali dulu —
 `orc upgrade` Anda adalah versi sebelum v0.56.0 dan tidak bisa memasang dirinya
-sendiri:
-
-- **Langkah 1 - lepaskan perintahnya dari paket lama:** `npm uninstall -g orc`
-- **Langkah 2 - pasang paket yang sekarang:** `npm i -g @azure-id/orc`
-- **Langkah 3 - terapkan lagi ke proyek Anda:** `orc update`
-
-**Jangan pakai `npm i -g -f`.** Rinciannya ada di v0.56.0 di bawah.
-
-**Pengiriman ke opencode yang diam di tengah jalan dulu menghabiskan seluruh
-batas waktu lima belas menit lalu melapor `timeout`.** Kata itu bercerita
-tentang kesabaran ORC, dan terbaca sebagai batas waktu yang perlu dinaikkan.
-Padahal yang tertinggal adalah sebuah POSISI yang perlu dilanjutkan seseorang.
-
-- **`extra_stall_s` (bawaan 180, `0` mematikannya)** menghentikan pekerja luar
-  yang tidak menghasilkan apa pun selama itu. Direset oleh kemajuan yang
-  terlihat - aliran keluaran pekerja, keluaran kesalahannya, atau berkas yang
-  dideklarasikan berubah di disk - jadi ia tidak pernah memotong pekerja yang
-  sekadar lambat. Hanya untuk mesin `cli`.
-- **`stalled` adalah jenis kegagalannya sendiri dan boleh diulang**, jadi
-  `orc extra reconcile` membaca posisinya dan `extra_resume` melanjutkan dari
-  apa yang sudah ada di disk, bukan mengulang dari nol. Itulah cara ORC sendiri
-  mengetik `continue`. Sengaja tidak ada dorongan lewat stdin: `opencode run`
-  bukan sesi percakapan, jadi ketikan yang tak dibaca siapa pun hanyalah
-  perbaikan palsu.
-- **Setiap balasan mesin `cli` membawa `timeline`** - bita pertama, kemajuan
-  terakhir, jeda diam terpanjang, dan kedua batas waktunya - baik saat berhasil
-  maupun gagal.
-- **`orc extra health <profil> [--model <id>]`** menjawab "apakah model ini suka
-  diam", lewat pengawas yang SAMA dengan yang dipakai pengiriman sungguhan.
-  Model yang terdaftar belum tentu jalan, dan model yang jalan belum tentu
-  selesai.
-- **`extra_fallback_agent` (bawaan `band`)** menentukan siapa yang mengambil
-  alih tugasnya. `ask` berhenti dan menyodorkan pilihannya kepada Anda; nama
-  agen apa pun yang terpasang bisa dikunci. Saat `ask`, lane-nya tidak memilih
-  sendiri. Ia mengubah siapa, bukan skor, daftar berkas, atau syarat
-  penerimaannya. Tidak berlaku di `/orc-quick`.
-
-### v0.56.0 - satu penggantian nama memindahkan perintahnya, dan tak seorang pun bisa meraih perbaikannya _(27-08-2026)_
-
-**Paketnya pindah dari `orc` tanpa awalan ke `@azure-id/orc`, dan semua jalur
-pemutakhiran di lapangan mati bersamaan.** Kedua nama itu mendaftarkan perintah
-`orc` yang sama, dan npm tidak mau menautkannya untuk paket baru selama paket
-lama masih memilikinya - jadi tarball, spesifikasi `github:`, dan registry
-semuanya gagal dengan `EEXIST` yang sama pada berkas perintahnya. Itu bentrokan
-BERKAS, bukan masalah sumber, dan itulah sebabnya berganti sumber tidak pernah
-menolong.
-
-- **`orc upgrade` mengusir paket lama SEBELUM mencoba sumber mana pun.**
-  Diumumkan, tidak pernah diam-diam. Pengenalannya lewat KEPEMILIKAN - paket
-  yang tidak mendaftarkan perintah `orc` tidak pernah disentuh.
-- **Registry npm dicoba lebih dulu**, lalu tarball, lalu spesifikasi `github:`.
-- **`--force` tetap ada untuk satu kasus yang cocok saja** - berkas perintah
-  yatim yang tidak dimiliki paket mana pun.
-- **`orc doctor` melaporkan `legacy-global-package` dengan namanya.** Sengaja
-  tidak bisa diperbaiki `--fix`: jangkauan `--fix` hanya `.claude/` proyek ini.
-
-### v0.55.1 — ORC sudah ada di npm _(27-08-2026)_
-
-**ORC diterbitkan dengan nama [`@azure-id/orc`](https://www.npmjs.com/package/@azure-id/orc).**
-Pemasangan lewat GitHub tetap jalan, dan isi paketnya tidak berubah sama sekali —
-ini hanya soal cara memasangnya sekarang punya nama.
-
-- **`npm i -g @azure-id/orc`** untuk memasang, dan
-  **`npm i -g @azure-id/orc@latest`** untuk memperbarui. `orc upgrade` sudah
-  melakukan keduanya sejak dulu dan tetap begitu.
-- **Pemasangan dari GitHub sekarang jadi cadangan**, bukan cara utama — tetap ada
-  di bagian "Cara mulai", dilipat, untuk fork atau kalau Anda mengunci sebuah
-  branch.
-
-### v0.51.0 — alat yang sudah Anda punya, dan koneksi yang membuktikan dirinya _(22-08-2026)_
-
-**Koneksi `orc extra`.** Dua alat coding yang bisa diberi pekerjaan oleh ORC
-adalah **program di komputer Anda sendiri**, bukan situs — jadi sekarang keduanya
-jadi penyedia kelas satu, punya kotak hubungkan masing-masing, dan daftar model
-yang dibangun dari apa yang benar-benar bisa dijangkau akun Anda.
-
-- **Program bisa saja belum terpasang, dan panelnya mengatakan itu lebih dulu** —
-  empat keadaan, dihitung setiap kali, tidak pernah diingat. Alat yang belum
-  terpasang tidak diberi tombol yang tidak mungkin berhasil, dan `orc extra add`
-  menolak sambil menyebutkan perintah pemasangannya.
-- **ORC membuka terminal Anda sendiri dan menjalankan pemasangannya di sana.**
-  Bukan pekerjaan tersembunyi di latar — di dalamnya, permintaan hak
-  administrator, kegagalan izin, unduhan 80 MB, dan tunggu empat puluh detik
-  semuanya terlihat sama: *tidak terjadi apa-apa*. Perintahnya tampil sebelum
-  tombolnya, jendelanya milik Anda, dan **ORC tidak pernah meminta hak
-  administrator**. Tidak ada terminal yang bisa dibuka? Anda tetap dapat
-  perintahnya untuk disalin.
-- **Uji koneksi sekarang bertingkat**: programnya ada · cukup baru · sudah masuk
-  akun · model apa yang bisa dijangkau akun ini · dan, hanya kalau Anda minta,
-  **apakah pesan sungguhan benar-benar dijawab** — lengkap dengan waktu tempuh,
-  jawabannya, dan empat jenis hitungan token yang tidak pernah dicampur.
-- **Model yang terdaftar bukan berarti model yang bekerja.**
-  `orc extra models <nama> --test <id>` adalah satu-satunya cara membedakannya.
-- **Kedua alat lokal itu tidak memberi tahu model mana yang menjawab**, jadi ORC
-  menuliskan kalimat itu, bukan membiarkan kolom kosong.
-- **`extra_enabled` tidak bisa dinyalakan sebelum ada yang menjawab** — dulu ia
-  akan terbaca NYALA padahal artinya MATI.
-
-Sebelumnya: **v0.50.0 — pekerjaan yang berjalan di tempat lain** (`orc extra`),
-**v0.49.1 — dewan penilai, dan `--json` yang berhenti membuang isinya**,
-**v0.49.0 — dokumen itu folder, dan berkasnya hasil rakitan** (`/orc-doc`), dan
-**v0.48.1 — satu berkas untuk satu hal, dan dokumen yang bisa diselesaikan**.
-[Baca semuanya di daftar perubahan](CHANGELOG.md).
+sendiri. Rincian lengkapnya ada di blok PERHATIAN di bagian atas berkas ini.
+
+- **Langkah 1 — lepaskan perintahnya dari paket lama:** `npm uninstall -g orc`
+- **Langkah 2 — pasang paket yang sekarang:** `npm i -g @azure-id/orc`
+- **Langkah 3 — terapkan lagi ke proyek Anda:** `orc update`
+
+**Tidak ada pengaturan yang berubah artinya, dan tidak ada perintah yang diganti
+namanya.** Tiga hal benar-benar mengubah perilaku; sisanya adalah ORC yang
+akhirnya membaca muatannya sendiri dengan cara yang selama ini ia ajarkan kepada
+Anda untuk membaca kode Anda.
+
+- **Tabel skor ke model berakhir di `opus-5-low [65,90)` · `opus-5-med
+  [90,100]`.** Dua dari enam pita kini ingin sesi utama Opus 5, dari sebelumnya
+  satu dari delapan.
+- **Pekerja luar yang macet dua kali dalam satu run akan menepi** untuk sisa run
+  itu. Dua jam, tidak pernah digabung; ia tidak menulis pengukuran baru dan tidak
+  pernah menulis konfigurasi Anda; sebuah promote adalah tanda batas, bukan
+  membisukan, dan perlu alasan.
+- **`orc diy init` kini berbawaan `opus-5-high`.** Bawaan yang lama diam-diam
+  meruntuhkan sepertiga atas tangga Anda ke satu agen sebelum Anda memilih
+  apa pun.
+
+Bagian strukturalnya - **konfigurasi, fase, dan panggilan berhenti jadi prosa**:
+
+- **`orc lane config <lane>`** menjawab konfigurasi sebuah lane sudah menjadi
+  apa, dengan setiap bayangan sudah dikalimatkan. Sebuah lane tidak pernah lagi
+  menggabungkan berkas konfigurasi sendiri, dan **peringkat di bawah peringkat
+  yang sudah menjawab tidak dibaca sama sekali**.
+- **`orc lane phases <lane>`** dan **`orc lane calls --all`** melakukan hal yang
+  sama untuk pustaka fase bersama dan katalog panggilan CLI - masing-masing satu
+  salinan kanonis, dari sebelumnya 14-59 pengulangan per panggilan.
+- **`orc ui` menampilkan semuanya**: tangga peringkat yang menunjukkan
+  pengaturan mana yang MENJAWAB, lane mana yang membaca tiap kunci, panel Lane
+  baru, dan baris penurunan di Extra ▸ Recovery beserta tombol Promote.
+  `orc doctor` mendapat `lane-keys-drifted`.
+
+**Ini tidak membuat muatannya lebih kecil** - 208 berkas menjadi 291, 26.507
+baris menjadi 33.204. Sebagian besar gelombang terukur sebagai kebenaran, bukan
+penghapusan duplikasi. Yang berubah adalah kini ada satu tempat untuk memperbaiki
+masing-masing hal ini, dan ada lint yang gagal saat sebuah salinan tumbuh
+kembali. Dua penghapusan yang direncanakan **diukur lalu ditolak**, dan flake
+yang membuat empat gelombang gagal akhirnya didiagnosis - dengan catatan jujur
+bahwa tiga run hijau adalah gerbangnya, bukan buktinya.
+
+**Entri lengkap: [CHANGELOG.md](CHANGELOG.md).**
 
 ---