I think a lot of people aren't using Grok Bot to its full potential, because knowing how to automate things is actually quite a difficult skill to learn. I wrote this post to break down how I approach automating anything with Grok Bot!
꼬뜨 retweeted
앱 개발과 게임 개발을 병행하고 있다.
보통 앱은 5개~10개 정도를 병렬로 작업하고
게임은 3~5개를 병렬로 작업하고 있다.
게임도 앱과 만드는 방식이 크게 다르지 않은것 같다.
MVP로 최대한 빠르고 핵심만 뽑고(앱과 다른점이 있다면 재미 요소가 하나라도 있는지 또 하고싶은 느낌이 드는지를 직접 플레이하여 꼼꼼히 체크한다)
스토어에 던진다.
그리고 열심히 광고를 돌려본다.
반응이 있으면 결제가 생각보다 빨리온다. 그럼 그점을 캐치하여 더 고도화 작업을 진행한다. 각종 도파민 세트를 기획하고 게임을 고도화 한다.
그리고 가장 앱과 다르게 운영하는 큰 부분이 있는데
pstack v0.15.43 更新情報
モデルの割り当てが Opus 5.5 中心で、どのロールもEffort が xhigh を上限にするようになりました!
さらに、/poteto-mode で起動する全エージェントは最初に worker.md(エージェント運用のルール) を読むようになりました!
それ以外では、Playbookが1件削除・各種Skillの縮小化・改善が行われました
■ New
・worker.md:/poteto-mode が起動したエージェント(Task のサブエージェントと Project のワーカー)が最初に読むルール集が追加
- 依頼文に書かれた合格ライン(基準)を、唯一の合格の条件として扱う。それより厳しいチェック、ゲート、承認、レビューのラウンドを、エージェント自身の判断で加えない
- 合格ラインがなければ、元のブランチより悪くならず、変更自体のテストが通ることを合格ラインとする
- 止まるのは、今のコミットで実行が失敗するときか、ユーザーが踏むバグがあるときだけで、元々ある問題は報告だけする
- 人に質問するツールは使わない。取り消せる選択は自分で決めて、何をなぜ選んだかを書く(デプロイ・データ削除・共有ブランチへの force-push・外部への連絡のときだけ止まる)
- PR は draft にせず ready で開く。タイトルは type(scope): subject、本文は40行以内で Why / What changed / Scope / Tradeoffs / Blast Radius / Verification の見出し
- 証拠のログ、実行の履歴、長い表は、ファイルかPRのコメントに置き、PRの説明からリンクする
- 自分がエージェントを起動するときも、依頼文の1行目で worker.md を読ませる
■ Improved
・モデル割り当て(/poteto-mode・/setup-pstack)が更新
- コード・文章・判断:claude-opus-5-5-xhigh-fast
- 些細な機械的編集・swarm workers・検証:claude-opus-5-5-high-fast
- パネル(arena runners・interrogate reviewers):claude-fable-5-2-high / claude-opus-5-5-high-fast / gemini-4-carbon-high
- arena の判定役の候補:claude-opus-5-5-high-fast / gemini-4-carbon-high(Fable は Opus と同じ提供元なので外れた)
- どのロールも xhigh を超えない。いちばん難しい作業でも1段階上げて xhigh まで。パネルの席は書かれた Level のまま
- Opus 5.5 は常に -fast。「急ぎのときだけ fast」の決まりはなくなった
- Fable 5.2 は、パネル以外では、ユーザーが名指ししたときか、同じ依頼で Opus 5.5 xhigh が2回チェックに落ちたときだけ。難しさだけでは使わない
- poteto-agent は claude-opus-5-5-xhigh-fast に固定。inherit-parent / auto のロールは generalPurpose で起動
- セカンドオピニオンと /show-me-your-work の点検役は、別の提供元のモデルにを設定するように
・/setup-pstack
- 予算は large / medium / small の3つ(xhigh > high > medium > low)。unlimited は large に移行
- hardest tasks・architect runners・reflect tooling・reflect judgment, divergent, synthesizer を廃止して、9ロールに
・/poteto-mode
- 大きい・横断的・放置して任せる仕事や、合う Playbook がない仕事は /figure-it-out へ
- 返答の書き方は1段落に
・/architect:変更が触るすべての仕組みをたどり、2つの形を描いて、状態の持ち主が1つ・やり方が1つ・呼び出し側が知ることが最も少ない形を選ぶ。回避策が繰り返したら作り直す。大きい設計や意見が割れる設計は /arena に回す
・/arena:候補のモデルが受け付けられないときは、同じ系統の近いスラッグを使うか席を外す(候補は最低2つ)。判定役は別の提供元の最初のモデルで、Fable は Claude として数える
・/reflect:サブエージェントを使わず、自分で会話を読む。半年後も通用し、次のエージェントの行動を変え、そのセッションで使った Skill に向けられる学びだけを残す。Accepted / Rejected / Backlog を見せて、承認を待ってから編集
・Claude Code 版:poteto-agent を起動するたびに model と effort を渡す。claude-fable-5-2-high は fable high に対応
■ その他の更新情報(削られた記述・内容)
・/poteto-mode:Playbook のステップを todolist ツールに写してから始める。飛ばした手順を skip: <理由> で残すのはやめた
・Opening a PR:3段落に。worktree、小さいコミット、/deslop、タイトルの形、本文の6つの見出し、ベースブランチで積む、ready で開く
・Babysit:6つの決まりに。モードは drive / threads-only / check。rebase・retarget・force-push はしない
・Multi-phase plan:5つの手順に。実行は運用者の「go」で始める
・principle 14個:本文を短縮(ルールとその確かめ方は残す)
・/unslop:本文を短縮。ルール番号はそのまま(廃止したルールは欠番)
・/technical-writing:4つの層を短い形に
・/poteto-help:Orchestrate、/architect の checkpoint、skip の説明を削除
・Orchestrate プレイブック(Playbook は22件に)
・/poteto-mode の Comments 節(/no-comments はレビュー前に今も実行)
・/setup-pstack の unlimited 予算と、hardest tasks・architect runners・reflect の2ロール
・/architect の参照ファイル3つ(設計の危険信号、根拠のテンプレート、候補への指示文)と、/reflect の参照ファイル4つ(3つのレビュー役とまとめ役)
・/architect の「with checkpoint」
・check-plan.mjs とテスト
꼬뜨 retweeted
Claude writes better code when it can verify its work. Checks that used to be expensive, like tests, benchmarks, and end-to-end runs, are now cheap.
Here’s how the Claude Code team thinks about verification:
claude.dev/blog/how-the-clau…
Replying to @cradrrat
We have dumped JAX/SGLang and rewritten the entire RL and inference stack from scratch in C/C++, which took some extra time.
Training (aka pre-training) is done. Lot of cleanup was needed on the SFT data to get that mostly right. Now, 4.8 is in RL and the evals are starting to climb. We will do a soft rollout to Bot users later this week and see how it goes.
Grok 4.8 will bear no resemblance to 4.7 (thankfully). It is not yet incredible, but, this time, the foundations are built right, so I expect it to evolve to be something special.
꼬뜨 retweeted
Anthropic has published a new guide to prompting Claude Fable 5.2.
Here are some of the key principles to consider when using it:
1. Effort now does a lot of the work - at high and above, Fable 5.2 thinks before almost every reply. The default is high but for every day coding and knowledge work, medium is fine.
2. It sometimes ends a turn before the work is done, especially at low and medium effort. Anthropic includes a block of text for your system prompt that roughly halved unfinished sessions in its testing.
3. It pays closer attention to who owns the code its working on. Adding “You are the owner and long-term maintainer of this codebase” meant it consolidated duplicated code in about 75% of runs, against under 10% without it.
4. It treats any instruction with “never” in it as absolute, so old cautions in memory files and notes can pile up and make it ask for approvals it doesn’t need. Anthropic suggests using the phrase "ask for approval before deleting files" instead of "never delete files"
Here’s the full guide:
platform.claude.com/docs/en/…
quick fixes considered harmful: you might be making your agents write worse code when you ask for the smallest or quickest possible change
what seems like a harmless bandaid may lead to that pattern being copied by the next agent. before you know it that pattern has spread everywhere
we used to write bandaid fixes by hand because we rarely had the time to do the right thing. agents change this calculus. sometimes you want it to fix the root cause instead even if it’s more invasive
Improvements to Grok @Bot efficiency
pstack の考え方はAI使う上では超汎用的に使える!
これは日本の pstack の第一人者として断言します!
だから、1つチャンクを上げて、思想・哲学・考え方を理解すれば、脳内 pstack として、どこでも持ち運びできるAI時代に最高の Skill Stack が手に入る
代表的なところでいうと以下が挙げられるけど、AI時代どれもかなり重宝させるものだろう
- 厳格で質の高い仕事の進め方
- エージェントが判断や迷った時に適用する原則
- 1つの /poteto-modeという Playbook の Router
- 推論の質を落とさないコンテキストウィンドウの運用管理
- エージェントに自分の仕事を確かめさせる検証 Skill と アプリの知識を与える Feature Map
- 過去を決定論に押し上げる /correct
- 過去を学ぶに変える /reflect
- エージェントの理解を引き出す方法
- 良い依頼文の書き方
応用の具体例は Ubie エージェント基盤もそうだし、Grok Bot でも活用できる
たとえば、Bot を組織的に使う場合、1つの Router(xx-mode)のようなものがいて、それで各厳格な手順を実施する個別の Bot がいるというような構成を作れる
個別の Bot に対しても同様に、依頼の種類に応じて Playbook を与えられるし、特定の条件下でどうすべきかの Principle も定義できる
Bot の勤怠管理の Bot なら、メンバー情報から部署を把握して、部署・役職ごとの Playbook を当てにしたり、コンテキストから何かしらの申請・承認があったと判明した場合の Principle を定義するなどができるだろう
これがProjectごとのハーネスや個別Skillでも同じような考え方が通用するし、実際それでかなり上手くいく
[Software Factoryの先へ:全社の仕事をAI-NativeにするUbieの内製クラウドエージェント基盤](zenn.dev/ubie_dev/articles/4…)
認証情報をサンドボックスの外に置くゲート型はもうデファクトスタンダードだなぁ。Ubie も pstack ベースのスキルを採用しているのはちょっと意外だった
꼬뜨 retweeted
Which Claude model should you build with?
Our new guide breaks down when to use Fable 5.2, Opus 5.5, Sonnet or Haiku. It also covers when to adjust effort instead of switching models, and how to save cost by putting subagents on smaller models.
claude.dev/blog/which-claude…
꼬뜨 retweeted
Design tip:
If your mobile app feels off but you're not sure why, it's probably the spacing.
Just did a round table review with the Grok @Bot engineering team.
Fixes are being implemented overnight and you should notice a dramatic improvement by tomorrow. If not, reply to this post and we will address any remaining issues right away.
꼬뜨 retweeted
You can now build live apps in Cursor and share them with a link.
Agents have made it easier to write code, but most apps never make it off localhost. Describe the tool you want - Cursor will build it, set up a database, run deterministic and agentic security checks, and deploy.
hot take: you don't need a software factory, and i don't love the term. it's just marketing
all you need is good software engineering and investing your tokens in the right places. the more deterministic the better (and token efficient)
while i'm quite bullish on formal methods, i think it's worth tempering expectations and doing your own research on it (disclaimer: i am not an expert in it). it cannot guarantee your code is bug free, only that it matches the spec under certain assumptions, and the specs come from what people want, not math. as long as we build software for humans, nothing can prove that your spec is right, only you can judge.
i created the /statechart skill in pstack primarily because it helps make the state machines implicit in your code more obvious and legible to you, so you can shrink the review surface and review the work at a higher level: does it actually match your intent and goal? i used this skill a ton while building my game tibo.dog, below you can see the statechart for story mode where i was able to find many bugs with it.
in that sense, i think model checking (eg TLA+ with TLC) is quite useful for understanding and checking properties of your system e2e, and you don't have to write tons of new verification code for it. the spec isn't your actual code so there's always a risk of drift, but so far i've had good experiences finding actual bugs with it. i tend to pair it with empirical verification (running the code) also to confirm it's real
other types of formal verification like proving theorems (eg Lean, Dafny), bounded model checking (eg CBMC, Kani) etc have their own tradeoffs. proofs cover every possible input, but only for the properties specified, and it takes a lot more code/work. bounded model checking can run on your actual code, which avoids drift, but it only checks up to a certain number of steps, so it's better at finding bugs than proving there aren't any
so i think in the near-medium term you'll still need to layer your codebase: constraints and guardrails (types, lint rules, api design, refactoring), empirical verification (tests, evals, real usage), and formal verification for the parts that really need it. none of these catch everything on their own, but layered together, they catch a lot more and shrink what's left for you to review
but also maybe in the limit, the bitter lesson will prevail and flawless code will just be printed out perfectly every time with no bugs. even in that scenario, /statechart still seems very helpful in confirming that the agent understood your intent correctly. give it a try if you haven't already!
github.com/cursor/plugins/tr…



























