dietrichgebert

[Plugins] See your GO and other usage live, auto-update your plugins, get task-done notification on your phone and give OpenCode a cup of coffee to keep it awake :D (check the video to see them in action) I've just updated my opencode-plugins suite with some other useful plugins I've published: as usual they work both in Desktop and CLI: oc-budget: a live indicator in the session header with the Go percentage and a ring. Click to open a popover showing the workspace it is reading, the three windows with their resets, and the balance on each of the four providers it covers: OpenCode Go, the Console credits, DeepSeek, OpenRouter. The CLI has the same thing in the sidebar, in the footer and a /budget command. It also marks the hours when a tariff doubles (peak hours). oc-plugin-updates: a pill above the composer that appears only when a plugin is behind. Which one, which version, one click, no restart. Pinned plugins are left alone on purpose. oc-notify: a push notification to your phone when a turn finishes while you are away from the computer. It goes through ntfy, whose app you install once. It has a settings page in OpenCode settings section. The others previously published: - oc-models-pro: every model with price in and out, context, intelligence, tok/s and the credit a request burns. Sorted by what it costs, one click to switch. The one I use most. - oc-git: branch, added, changed, untracked, ahead and behind, in the sidebar and as a pill. - oc-coffee: a cup of real Italian espresso in the session header and in the CLI footer, lit only while the computer is kept awake, during real tasks, with four levels behind a click (macOS only). - oc-ponytail-widget: the ponytail level in the footer, and a pill to change it. - oc-session-search: keyword search over your past sessions, as a tool. To install: opencode plugin add @rizal72/oc-budget opencode plugin add @rizal72/oc-plugin-updates opencode plugin add @rizal72/oc-notify The other five are in the same repository, with their pages and READMEs: rizal72/opencode-plugins

Forget the usual Claude tips. What are the strange things you do that actually help? I feel like everyone whos used claude for a while ends up developing their own weirdly specific workflows and habits. And i think they probably sound ridiculous to other people, but we keep using them and they keep working I'm very curious what some of these are for other people, especially the less obvious ones I feel I do all the normal stuff like starting fresh chats for unrelated tasks, keeping my claude.md and documentation short, and a couple other little things. But i think i have a couple that are a little strange but make a big difference to me 1. I almost never use skills or typical plugins I find that their descriptions can add unnecessary context, and sometimes Claude ends up calling them when it really shouldn't. I find with the models these days and using voice transcription I dont need skills to enforce things, or give extra guidance, the models just know. I tend to keep things in a skills folder somewhat but just local to that project, and only for very specific things 2. Claude memory is OFF Not sure if this one is more common, but I find claude always makes insanely bad memories about finishing tasks/passes, documents terrible out of date old assumptions, states parts of the code as "working dont check" or other horrible things that make future agents more and more stupid and lazy. Claude would just be so unpredictable and behave very differently between projects, and turning it off made things way better 3. Token saving tools I love maxing out my efficiency and just optimizing for random numbers (just a personal thing i think). But it has gotten me into using a bunch of different tools to reduce my token usage per task, and basicaly make my usage last way longer. Opus 5.5 has been great for usage, but the difference between when I stop using them to when I am is kind of wild. Also another byproduct of reducing token usage is my agents also work like way faster, which is quite convenient since on higher tier plans the time to complete tasks feels like more of a bottleneck than the usage since I have started using these tools. I have been using Graphify ( Graphify-Labs/graphify ) and StateSync ( statesync.net ) quite extensively the last while and have been trying out Ponytail ( dietrichgebert/ponytail ) I have really enjoyed graphify and statesync since they dont seem to change how they AI feels to work with at all or how good of a job it does, and they just make it so much cheaper and faster. Ponytail is seeming to make a difference, at least with the number of random tests, but it changes how the AI responds and interprets my instructions, and I haven't worked out if it is worth the tradeoff yet. 4. Prompt writing I feel like this is somewhat of a 'dying art' if you could call it that. I think with voice transcription being able to just give bucketloads of detail, when paired with the insane understanding of claude these days removes the need for well crafted prompts alot of the time. Although, when i do have something important, or a new feature set I want to work on, I often open up a ChatGPT chat on the web, give it my prompt writing skill, and then just transcribe a crazy monologue, then answer a couple questions, and come out with a really good prompt that I touch up before sending off. The reason I think this is valuable still is because for larger work it is hard to think of everything up front, and to change things later is relatively expensive in terms of tokens compared to building it in first time. And having those questions helps me think of more detail than i would normaly, and forces me to question what I am asking for and if it is actualy the best way. Anyway, those are mine. I'm curious what other people are doing

Forget the usual Claude tips. What are the strange things you do that actually help? I feel like everyone whos used claude for a while ends up developing their own weirdly specific workflows and habits. And i think they probably sound ridiculous to other people, but we keep using them and they keep working I'm very curious what some of these are for other people, especially the less obvious ones I feel I do all the normal stuff like starting fresh chats for unrelated tasks, keeping my claude.md and documentation short, and a couple other little things. But i think i have a couple that are a little strange but make a big difference to me 1. I almost never use skills or typical plugins I find that their descriptions can add unnecessary context, and sometimes Claude ends up calling them when it really shouldn't. I find with the models these days and using voice transcription I dont need skills to enforce things, or give extra guidance, the models just know. I tend to keep things in a skills folder somewhat but just local to that project, and only for very specific things 2. Claude memory is OFF Not sure if this one is more common, but I find claude always makes insanely bad memories about finishing tasks/passes, documents terrible out of date old assumptions, states parts of the code as "working dont check" or other horrible things that make future agents more and more stupid and lazy. Claude would just be so unpredictable and behave very differently between projects, and turning it off made things way better 3. Token saving tools I love maxing out my efficiency and just optimizing for random numbers (just a personal thing i think). But it has gotten me into using a bunch of different tools to reduce my token usage per task, and basicaly make my usage last way longer. Opus 5.5 has been great for usage, but the difference between when I stop using them to when I am is kind of wild. Also another byproduct of reducing token usage is my agents also work like way faster, which is quite convenient since on higher tier plans the time to complete tasks feels like more of a bottleneck than the usage since I have started using these tools. I have been using Graphify ( Graphify-Labs/graphify ) and StateSync ( statesync.net ) quite extensively the last while and have been trying out Ponytail ( dietrichgebert/ponytail ) I have really enjoyed graphify and statesync since they dont seem to change how they AI feels to work with at all or how good of a job it does, and they just make it so much cheaper and faster. Ponytail is seeming to make a difference, at least with the number of random tests, but it changes how the AI responds and interprets my instructions, and I haven't worked out if it is worth the tradeoff yet. 4. Prompt writing I feel like this is somewhat of a 'dying art' if you could call it that. I think with voice transcription being able to just give bucketloads of detail, when paired with the insane understanding of claude these days removes the need for well crafted prompts alot of the time. Although, when i do have something important, or a new feature set I want to work on, I often open up a ChatGPT chat on the web, give it my prompt writing skill, and then just transcribe a crazy monologue, then answer a couple questions, and come out with a really good prompt that I touch up before sending off. The reason I think this is valuable still is because for larger work it is hard to think of everything up front, and to change things later is relatively expensive in terms of tokens compared to building it in first time. And having those questions helps me think of more detail than i would normaly, and forces me to question what I am asking for and if it is actualy the best way. Anyway, those are mine. I'm curious what other people are doing

Dev OS v2: gates, a learning loop, and agents that check agents In v1 I treated ~/.claude as an operating system around the model: a kernel (one rulebook), a scheduler (lanes), interrupts (hooks), a package manager (pinned skills). I asked you to run the starter kit for a week and tell me what broke. I ran it too, every day. 213 commits later, this is the v2 release, and its main change fits on one line: > A rule the agent reads is a request. A rule something runs is a gate. Both commenters on v1 had said the same thing. Here is what that changed, in order of how much it helped. The gates, hooks, agents and tests below ship in the kit (135 checks, bash tests/run.sh); the rest is how I run my own setup. The kit works with any agent that reads AGENTS.md (Codex, Cursor, Copilot), not only Claude Code. 1. One rulebook, every agent The kernel is the same: one CLAUDE.md whose first section says it overrides every skill, plugin and repo file. What's new is that AGENTS.md is a symlink to it. ln -s ~/.claude/CLAUDE.md ~/.codex/AGENTS.md # Codex reads ~/.codex/AGENTS.md, then one per folder, root → cwd I checked how each tool loads it on 2026-11-01: - Codex concatenates AGENTS.md files up to 32 KiB by default. A long global rulebook steals room from repo rules. Mine is capped at 136 lines by a test. - Claude Code reads AGENTS.md only in a repo with no CLAUDE.md, so CLAUDE.md stays the source. - @lessons.md imports are Claude-only. Through the link, Codex gets the rulebook but not its imports. 2. Gates beat reminders A test gate in plain git. v1's pre-commit hook only guarded ~/.claude itself. v2 installs one per project: echo 'npm test' > .dev-os-test && ~/.claude/bin/gate-install The hook it writes (condensed from bin/gate-install): TOP=$(git rev-parse --show-toplevel) unset GIT_INDEX_FILE GIT_DIR GIT_WORK_TREE GIT_PREFIX CMD=$(cat "$TOP/.dev-os-test" 2>/dev/null) if ! printf '%s' "$CMD" | grep -q '[^[:space:]]'; then echo "dev-os test gate: .dev-os-test is missing or empty; commit refused" >&2; exit 1 fi (cd "$TOP" && bash -o pipefail -c "$CMD") || { echo "dev-os test gate: commit refused" >&2; exit 1; } Every line there is a bug I hit: - pipefail: npm test | tail -5 exits with tail's 0, so a red suite commits. - Empty means refuse: a missing or empty test command fails loudly instead of passing quietly. - unset GIT_INDEX_FILE: git runs pre-commit with a temporary index during commit -a. A test suite that runs git add in a temp repo writes into your commit and breaks it with "Error building trees". - No global hooks: the installer refuses when a global core.hooksPath would turn one repo's gate into every repo's gate. Permissions the agent can't talk past. permissions.ask rules still prompt under --dangerously-skip-permissions, and compound commands are split before matching: "permissions": { "ask": ["Bash(git push:*)", "Bash(gh pr create:*)"] } One catch: rules match the command text, so git -C . push slips past Bash(git push:*). I add Bash(git -C * push *) in my own settings. Keep branch protection on the remote as the backstop. A Stop hook for the cheap way out. v1's Stop hook sent the agent back when source changed with no test. v2 adds one more case: a weakened test. That means fewer assertions, a new .skip/.only, or a deleted test file. ASSERT = re.compile(r"\bassert\w*\s*\(|\bexpect\s*\(|->assert\w*\(|\$this->assert|^[-+]\s*assert(_\w+)?\s") SKIP = re.compile(r"markTestSkipped|markTestIncomplete|->skip\(|\.(skip|only|fixme)\s*\(|\bx(it|describe|test)\s*\(") if removed_asserts > added_asserts or any(SKIP.search(l) for l in added): weakened.append(path) # → {"decision": "block", "reason": "TEST WEAKENED: …"} Codex hooks use the same Stop → {"decision":"block"} shape. 3. It learns by itself In v1, "add a line to lessons.md" depended on me remembering to do it. Now the setup logs its own mistakes: prompt starts "no, …" / "I said …" ─┐ stop-check or test gate blocks ─┼─▶ state/learn-<session>.jsonl │ Stop: learn.sh → any entry newer than the last `reflect`? → block once: "run reflect" reflect routes each mistake to the lowest layer that would have caught it: first a test or hook, then one line in the skill that owns the job, and only last a line in lessons.md. lessons.md has a 60-line budget. When it fills up, meditate audits the rules for stale, duplicate and contradicting lines, and deletes nothing without my yes. 59 of the 213 commits are Lesson: commits, and many of them removed a lesson by turning it into a check. The loop gets tested like everything else. A cold review caught the correction detector flagging "No tests exist for the parser yet" as a correction. Now no counts only when followed by punctuation, and only the first line of a prompt is read. 4. Agents check agents One agent reviewing its own diff is a student grading their own exam. The kit ships two read-only agent shapes: - scout: one angle per scout (callers, tests, history, config). Each returns file:line facts and what it couldn't check. - review-lens: Skeptic always, Architect at 50+ changed lines, Minimalist at 200+. Then one verifier per finding that starts from "false positive". Two skills run them. pre-merge-review is relay step 7 on your own branch: lenses sized by the diff, a verifier per finding, one saved report. post-pr-review is for someone else's PR: each finding must reproduce on both the base and the PR branch before it goes in a draft, and nothing is posted without a yes. /code-review stays the per-slice check inside build-check, and it never posts on other people's PRs. Every subagent brief has the same fields: goal, scope, what's known, what's been ruled out, the exact ask, the output format, what's out of scope. Parallel builders get one git worktree each, and only for plan slices that share no files. How well does this work? On the v2 kit itself, which already passed all its own tests: | Pass | Found | |---|---| | /code-review | 9 (6 fixed, 3 left with reasons) | | 3 lenses, each reproducing its findings in a temp dir | 14 more | | a cold reviewer that saw none of the above | 4 more | Every behaviour fix got a red test first. 5. Prove it by running it 9. A bug or security claim counts only when reproduced by running it. No repro, no fix. For bugs, the first job is a red-capable feedback loop, in the order from Matt Pocock's diagnosing-bugs: failing test → curl script → CLI snapshot diff → headless browser → replayed payload → harness → fuzz → git bisect run. Hypotheses come after. I also break every gate on purpose before I trust it. That caught a test that passed before the code existed. My helper matched substrings, so rc=127 ("command not found") satisfied a check for rc=1. 6. Project memory: graph first Rule 8: before grep, ask a code graph "who calls this, and what breaks if I change it?" | Tool | Used for | |---|---| | graphify query / explain / path / merge-graphs | Finding code, explaining a node, call paths, one graph across repos | | code-review-graph get_impact_radius_tool | Who else a fix reaches (the no-regressions rule) | | get_review_context_tool, get_affected_flows_tool | Review context; which E2E flows a diff touches | The scouts, review lenses and review-reply step all start there. The graph holds structure. What the project taught lives in lessons.md and in plan folders kept outside the repo. 7. Stay lean - A context budget test: rulebook plus skill descriptions, measured in bytes, against a baseline. It went from 59,287 to 46,927 while the setup grew. - Routing is tested: one rulebook sentence ("call the owner skill before any other tool") took my routing eval from 0-1 to 7-8 of 11 prompts. - Effort is a budget too: "effortLevel": "medium" by default, in every modelSettings entry as well, since a per-model value wins. High effort lives in skill frontmatter only where a wrong answer is expensive: grilling, writing and checking the plan, whole-plugin bug hunts. Hooks can read the effort level but not set it, so serious cases are handled by a rule: payment or refund, data loss, auth or a security finding mean the verifiers are spawned with effort: "high". - Smaller output: RTK trims git/gh/test output before the agent reads it. Compressed output can hide the one line you need, so "never report a result you couldn't see" is a rule. The userland Most of my 103 skills are other people's work: 39 are pinned to an upstream commit in skills.lock, with local edits in overlays that a sync script re-applies. - ponytail by Dietrich Gebert: the smallest-change ladder (YAGNI → reuse → stdlib → platform → installed dep → one line). It's rule 3 in the kit. - Matt Pocock's skills: tdd, grill-me, handoff, domain-modeling. - obra/superpowers: brainstorming and systematic-debugging. - Trail of Bits: fp-check, variant-analysis and differential-review, for security claims. - e2e-reviewer: catches E2E tests that pass while the feature is broken. Skills use the Agent Skills format. Codex reads them from .agents/skills, and Claude Code from .claude/skills, so one folder with two symlinks serves both. Upgrade or start New: follow the kit's README "Start here". The core (rulebook, test gate, plan template) takes about 10 minutes, and the Claude Code layer is optional. From v1: the README has an "Upgrading from v1" block. Nothing was renamed. Two hooks and two skills get replaced after a "Before v2" commit, so it's one git revert to undo. Same ask as v1: run it for a week and tell me what broke. And what's the one rule your agent keeps breaking? I'll tell you which gate I'd put behind it.

Is this entire repo satire? It's a bunch of text files in a repo with massive churn (all new version 5!), but somehow even the technical debt 'skill' (i.e. text file) has technical debt itself: github.com/DietrichGebert/ponytail/blob/main/skil…... > ...every ponytail shortcut is marked with a shortcut: comment ... (older code may say ponytail:).

He doesnt need that when he has Ponytail DietrichGebert/ponytail Right? ...

This Free GitHub Repo Makes Your AI Write 10 Lines, Not 335 Ponytail is a free GitHub repo with over 170,000 stars that makes your AI agent think like the laziest senior dev in the room: reuse what's already there and write far less code, without cutting validation, error handling or security. Works with Claude Code, Codex, Cursor and more. Get it free here: DietrichGebert/ponytail Follow for more AI tips. #github #claudecode #coding #aitools #opensource #shorts [ar] هذا المستودع المجاني على GitHub يجعل ذكاءك الاصطناعي يكتب 10 أسطر برمجية بدلاً من 335 بونيتيل هو مستودع مجاني على GitHub، حائز على أكثر من 170,000 نجمة، يجعل وكيل الذكاء الاصطناعي الخاص بك يفكر كأكثر مطور برامج متمرس كسلاً: إعادة استخدام ما هو موجود بالفعل وكتابة كود أقل بكثير، دون المساس بالتحقق من صحة البيانات أو معالجة الأخطاء أو الأمان. يعمل مع Claude Code وCodex وCursor وغيرها. حمّله مجانًا من هنا: DietrichGebert/ponytail تابعنا للمزيد من نصائح الذكاء الاصطناعي. #github #claudecode #coding #aitools #opensource #shorts [bn] এই ফ্রি GitHub রিপো আপনার AI-কে ৩৩৫ লাইনের বদলে ১০ লাইনে কোড লিখতে সাহায্য করে Ponytail হলো একটি ফ্রি গিটহাব রিপো, যার ১৭০,০০০-এরও বেশি স্টার রয়েছে। এটি আপনার AI এজেন্টকে এমনভাবে ভাবতে শেখায় যেন সে সবচেয়ে অলস সিনিয়র ডেভেলপার: আগে থেকে যা আছে তা পুনরায় ব্যবহার করে এবং অনেক কম কোড লেখে, অথচ ভ্যালিডেশন, এরর হ্যান্ডলিং বা সিকিউরিটিতে কোনো ছাড় দেয় না। এটি Claude Code, Codex, Cursor এবং আরও অনেক কিছুর সাথে কাজ করে। এটি এখানে বিনামূল্যে পান: DietrichGebert/ponytail আরও AI টিপসের জন্য ফলো করুন। #github #claudecode #coding #aitools #opensource #shorts [de-DE] Dieses kostenlose GitHub-Repo lässt deine KI 10 Zeilen Code schreiben statt 335 Ponytail ist ein kostenloses GitHub-Repository mit über 170.000 Sternen, das Ihren KI-Agenten dazu bringt, wie der faulste Senior-Entwickler im Raum zu denken: Vorhandenen Code wiederverwenden und deutlich weniger Code schreiben, ohne dabei auf Validierung, Fehlerbehandlung oder Sicherheit zu verzichten. Funktioniert mit Claude Code, Codex, Cursor und mehr. Kostenlos hier herunterladen: DietrichGebert/ponytail Folgen Sie uns für weitere KI-Tipps. #github #claudecode #coding #aitools #opensource #shorts [es-US] Este repositorio de GitHub gratuito hace que tu IA escriba 10 líneas en lugar de 335 Ponytail es un repositorio gratuito de GitHub con más de 170.000 estrellas que permite que tu agente de IA piense como el desarrollador senior más perezoso de la sala: reutiliza lo que ya existe y escribe mucho menos código, sin sacrificar la validación, el manejo de errores ni la seguridad. Funciona con Claude Code, Codex, Cursor y más. Descárgalo gratis aquí: DietrichGebert/ponytail Sígueme para más consejos sobre IA. #github #claudecode #programación #herramientasdeIA #códigoabierto #shorts [fr-FR] Ce dépôt GitHub gratuit fait écrire 10 lignes de code à votre IA au lieu de 335 Ponytail est un dépôt GitHub gratuit, plébiscité par plus de 170 000 utilisateurs, qui permet à votre agent IA de penser comme le développeur senior le plus paresseux : réutiliser le code existant et écrire beaucoup moins de code, sans compromettre la validation, la gestion des erreurs ni la sécurité. Compatible avec Claude Code, Codex, Cursor et bien d'autres. Téléchargez-le gratuitement ici : DietrichGebert/ponytail Suivez-nous pour plus de conseils en IA. #github #claudecode #coding #aitools #opensource #shorts [hi] यह फ्री GitHub रिपो आपके AI से 335 की जगह केवल 10 लाइन लिखवाती है Ponytail एक मुफ़्त GitHub रिपॉज़िटरी है जिसे 170,000 से ज़्यादा स्टार मिल चुके हैं। यह आपके AI एजेंट को कमरे में मौजूद सबसे आलसी सीनियर डेवलपर की तरह सोचने में सक्षम बनाता है: पहले से मौजूद चीज़ों का पुन: उपयोग करें और बहुत कम कोड लिखें, बिना वैलिडेशन, एरर हैंडलिंग या सुरक्षा में कोई कमी किए। यह Claude Code, Codex, Cursor और अन्य के साथ काम करता है। इसे यहाँ से मुफ़्त में प्राप्त करें: DietrichGebert/ponytail अधिक AI टिप्स के लिए फ़ॉलो करें। #github #claudecode #coding #aitools #opensource #shorts [id] Repo GitHub Gratis Ini Bikin AI Anda Menulis 10 Baris, Bukan 335 Ponytail adalah repositori GitHub gratis dengan lebih dari 170.000 bintang yang membuat agen AI Anda berpikir seperti pengembang senior paling malas di ruangan: menggunakan kembali apa yang sudah ada dan menulis kode jauh lebih sedikit, tanpa mengurangi validasi, penanganan kesalahan, atau keamanan. Bekerja dengan Claude Code, Codex, Cursor, dan banyak lagi. Dapatkan secara gratis di sini: DietrichGebert/ponytail Ikuti untuk tips AI lainnya. #github #claudecode #coding #aitools #opensource #shorts [it] Questo repository GitHub gratuito fa scrivere alla tua IA 10 righe anziché 335 Ponytail è un repository GitHub gratuito con oltre 170.000 stelle che permette al tuo agente IA di pensare come il più pigro degli sviluppatori senior: riutilizza ciò che è già presente e scrivi molto meno codice, senza compromettere la validazione, la gestione degli errori o la sicurezza. Funziona con Claude Code, Codex, Cursor e altri. Scaricalo gratuitamente qui: DietrichGebert/ponytail Seguici per altri consigli sull'IA. #github #claudecode #coding #aitools #opensource #shorts [iw] המאגר החינמי הזה ב-GitHub גורם ל-AI שלך לכתוב 10 שורות קוד במקום 335 Ponytail הוא מאגר חינמי של GitHub עם מעל 170,000 כוכבים שגורם לסוכן הבינה המלאכותית שלכם לחשוב כמו המפתח הבכיר והעצלן ביותר בחדר: השתמשו מחדש במה שכבר קיים וכתבו הרבה פחות קוד, מבלי לקצץ באימות, טיפול בשגיאות או באבטחה. עובד עם Claude Code, Codex, Cursor ועוד. הורידו אותו בחינם כאן: DietrichGebert/ponytail עקבו אחרינו לקבלת טיפים נוספים בנושא בינה מלאכותית. #github #claudecode #coding #aitools #opensource #shorts [ja] AIのコードを335行から10行に激減させる無料のGitHubリポジトリ Ponytailは、17万以上のスターを獲得している無料のGitHubリポジトリです。AIエージェントを、まるで部屋で一番怠惰なベテラン開発者のように考えさせます。既存のものを再利用し、検証、エラー処理、セキュリティを損なうことなく、コード量を大幅に削減します。Claude Code、Codex、Cursorなどと連携します。 こちらから無料で入手できます：DietrichGebert/ponytail AIに関するその他のヒントについては、フォローしてください。 #github #claudecode #coding #aitools #opensource #shorts [ko] AI가 335줄 대신 10줄만 쓰게 만드는 무료 GitHub 저장소 Ponytail은 17만 개 이상의 별을 받은 무료 GitHub 저장소로, AI 에이전트가 마치 가장 게으른 시니어 개발자처럼 생각하도록 도와줍니다. 기존 코드를 재사용하여 훨씬 적은 코드를 작성하면서도 유효성 검사, 오류 처리, 보안은 그대로 유지할 수 있습니다. Claude Code, Codex, Cursor 등 다양한 도구와 호환됩니다. 여기에서 무료로 다운로드하세요: DietrichGebert/ponytail 더 많은 AI 팁을 보려면 팔로우하세요. #github #claudecode #coding #aitools #opensource #shorts [ml] ഈ സൗജന്യ GitHub റിപ്പോ നിങ്ങളുടെ AI-യെക്കൊണ്ട് 335 വരികൾക്ക് പകരം 10 വരികൾ മാത്രം എഴുതിക്കുന്നു 170,000-ത്തിലധികം നക്ഷത്രങ്ങളുള്ള ഒരു സൗജന്യ GitHub ശേഖരമാണ് പോണിടെയിൽ, ഇത് നിങ്ങളുടെ AI ഏജന്റിനെ ഏറ്റവും മടിയനായ മുതിർന്ന ഡെവലപ്പറെപ്പോലെ ചിന്തിക്കാൻ പ്രേരിപ്പിക്കുന്നു: വാലിഡേഷൻ, പിശക് കൈകാര്യം ചെയ്യൽ അല്ലെങ്കിൽ സുരക്ഷ എന്നിവ കുറയ്ക്കാതെ, നിലവിലുള്ളത് വീണ്ടും ഉപയോഗിക്കുകയും വളരെ കുറച്ച് കോഡ് എഴുതുകയും ചെയ്യുക. ക്ലോഡ് കോഡ്, കോഡെക്സ്, കഴ്‌സർ എന്നിവയിലും മറ്റും പ്രവർത്തിക്കുന്നു. ഇവിടെ സൗജന്യമായി നേടൂ: DietrichGebert/ponytail കൂടുതൽ AI നുറുങ്ങുകൾക്കായി പിന്തുടരുക. #github #claudecode #coding #aitools #opensource #shorts [nl-NL] Deze gratis GitHub-repo laat je AI 10 regels code schrijven in plaats van 335 Ponytail is een gratis GitHub-repository met meer dan 170.000 sterren waarmee je AI-agent kan denken als de meest luie senior ontwikkelaar: hergebruik bestaande code en schrijf veel minder code, zonder in te leveren op validatie, foutafhandeling of beveiliging. Werkt met Claude Code, Codex, Cursor en meer. Download het hier gratis: DietrichGebert/ponytail Volg voor meer AI-tips. #github #claudecode #coding #aitools #opensource #shorts [pa] ਇਹ ਮੁਫਤ GitHub ਰਿਪੋ ਤੁਹਾਡੀ AI ਨੂੰ 335 ਦੀ ਬਜਾਏ ਸਿਰਫ 10 ਲਾਈਨਾਂ ਲਿਖਣ ਲਈ ਮਜਬੂਰ ਕਰਦੀ ਹੈ ਪੋਨੀਟੇਲ 170,000 ਤੋਂ ਵੱਧ ਸਟਾਰਾਂ ਵਾਲਾ ਇੱਕ ਮੁਫ਼ਤ GitHub ਰੈਪੋ ਹੈ ਜੋ ਤੁਹਾਡੇ AI ਏਜੰਟ ਨੂੰ ਕਮਰੇ ਵਿੱਚ ਸਭ ਤੋਂ ਆਲਸੀ ਸੀਨੀਅਰ ਡਿਵੈਲਪਰ ਵਾਂਗ ਸੋਚਣ ਲਈ ਮਜਬੂਰ ਕਰਦਾ ਹੈ: ਪਹਿਲਾਂ ਤੋਂ ਮੌਜੂਦ ਚੀਜ਼ਾਂ ਦੀ ਮੁੜ ਵਰਤੋਂ ਕਰੋ ਅਤੇ ਪ੍ਰਮਾਣਿਕਤਾ, ਗਲਤੀ ਪ੍ਰਬੰਧਨ ਜਾਂ ਸੁਰੱਖਿਆ ਨੂੰ ਕੱਟੇ ਬਿਨਾਂ ਬਹੁਤ ਘੱਟ ਕੋਡ ਲਿਖੋ। ਕਲਾਉਡ ਕੋਡ, ਕੋਡੈਕਸ, ਕਰਸਰ ਅਤੇ ਹੋਰ ਬਹੁਤ ਕੁਝ ਨਾਲ ਕੰਮ ਕਰਦਾ ਹੈ। ਇਸਨੂੰ ਇੱਥੇ ਮੁਫ਼ਤ ਪ੍ਰਾਪਤ ਕਰੋ: DietrichGebert/ponytail ਹੋਰ AI ਸੁਝਾਵਾਂ ਲਈ ਫਾਲੋ ਕਰੋ। #github #claudecode #coding #aitools #opensource #shorts [pl] Dzięki temu darmowemu repozytorium z GitHub Twoje AI pisze 10 linii kodu zamiast 335 Ponytail to darmowe repozytorium GitHub z ponad 170 000 gwiazdek, które sprawi, że Twój agent AI będzie myślał jak najbardziej leniwy, doświadczony programista: wykorzystaj to, co już istnieje, i napisz znacznie mniej kodu, bez rezygnowania z walidacji, obsługi błędów i zabezpieczeń. Działa z Claude Code, Codex, Cursor i innymi. Pobierz za darmo tutaj: DietrichGebert/ponytail Obserwuj, aby uzyskać więcej wskazówek dotyczących AI. #github #claudecode #coding #aitools #opensource #shorts [pt-BR] Este repositório gratuito do GitHub faz sua IA escrever 10 linhas, em vez de 335 Ponytail é um repositório gratuito no GitHub com mais de 170.000 estrelas que faz seu agente de IA pensar como o desenvolvedor sênior mais preguiçoso da sala: reutilizar o que já existe e escrever muito menos código, sem comprometer a validação, o tratamento de erros ou a segurança. Funciona com Claude Code, Codex, Cursor e muito mais. Baixe gratuitamente aqui: DietrichGebert/ponytail Siga para mais dicas de IA. #github #claudecode #coding #aitools #opensource #shorts [ru] Этот бесплатный репозиторий на GitHub заставляет ваш ИИ писать 10 строк кода вместо 335 Ponytail — это бесплатный репозиторий на GitHub с более чем 170 000 звёзд, который заставляет вашего ИИ-агента мыслить как самый ленивый старший разработчик в комнате: повторно использовать уже существующий код и писать гораздо меньше кода, не жертвуя валидацией, обработкой ошибок или безопасностью. Работает с Claude Code, Codex, Cursor и другими инструментами. Скачать бесплатно можно здесь: DietrichGebert/ponytail Подписывайтесь, чтобы получать больше советов по ИИ. #github #claudecode #coding #aitools #opensource #shorts [ta] உங்கள் AI-ஐ 335 வரிகளுக்குப் பதில் 10 வரிகள் மட்டும் எழுத வைக்கும் இந்த இலவச GitHub ரெப்போ! போனிடெயில் (Ponytail) என்பது 170,000-க்கும் மேற்பட்ட ஸ்டார்களைக் கொண்ட ஒரு இலவச கிட்ஹப் ரெப்போ (GitHub repo) ஆகும். இது உங்கள் AI ஏஜென்ட்டை, அறையில் உள்ள மிகவும் சோம்பேறியான மூத்த டெவலப்பரைப் போல சிந்திக்க வைக்கிறது: சரிபார்ப்பு, பிழை கையாளுதல் அல்லது பாதுகாப்பைக் குறைக்காமல், ஏற்கனவே உள்ளவற்றை மீண்டும் பயன்படுத்தி, மிகக் குறைவான குறியீட்டை எழுத வைக்கிறது. இது கிளாட் கோட் (Claude Code), கோடெக்ஸ் (Codex), கர்சர் (Cursor) மற்றும் பலவற்றுடன் செயல்படுகிறது. இங்கே இலவசமாகப் பெறுங்கள்: DietrichGebert/ponytail மேலும் AI குறிப்புகளுக்குப் பின்தொடரவும். #github #claudecode #coding #aitools #opensource #shorts [te] ఈ ఉచిత GitHub రిపో మీ AIని 335 లైన్లకు బదులుగా కేవలం 10 లైన్లు మాత్రమే రాసేలా చేస్తుంది పోనీటెయిల్ అనేది 170,000 కంటే ఎక్కువ స్టార్స్ ఉన్న ఒక ఉచిత గిట్‌హబ్ రిపో. ఇది మీ AI ఏజెంట్‌ను, ఆ గదిలోని అత్యంత సోమరి సీనియర్ డెవ్ లాగా ఆలోచించేలా చేస్తుంది: వాలిడేషన్, ఎర్రర్ హ్యాండ్లింగ్ లేదా సెక్యూరిటీలో ఎలాంటి కోత పెట్టకుండా, ఇప్పటికే ఉన్నవాటిని తిరిగి ఉపయోగించుకుంటూ చాలా తక్కువ కోడ్ రాసేలా చేస్తుంది. ఇది క్లాడ్ కోడ్, కోడెక్స్, కర్సర్ మరియు మరిన్నింటితో పనిచేస్తుంది. ఇక్కడ ఉచితంగా పొందండి: DietrichGebert/ponytail మరిన్ని AI చిట్కాల కోసం ఫాలో అవ్వండి. #github #claudecode #coding #aitools #opensource #shorts [uk] Цей безкоштовний репозиторій GitHub змушує ваш ШІ писати 10 рядків коду замість 335 Ponytail — це безкоштовний репозиторій на GitHub з понад 170 000 зірок, який змушує вашого ШІ-агента думати як найлінивішого старшого розробника в кімнаті: повторно використовуйте те, що вже є, і пишіть набагато менше коду, не порушуючи валідацію, обробку помилок чи безпеку. Працює з Claude Code, Codex, Cursor та іншими. Завантажте безкоштовно тут: DietrichGebert/ponytail Підпишіться, щоб отримати більше порад щодо ШІ. #github #claudecode #coding #aitools #opensource #shorts

This GitHub Repo Makes Claude Use 45% Fewer Tokens Ponytail makes your AI agent think like the laziest senior dev in the room: it writes only the code you actually need. Day 8 of Repos You Should Know 👇 🔗 DietrichGebert/ponytail Before writing code, it stops at the first option that works: skip it, reuse what's already in your repo, use the standard library or a platform feature, make it one line, and only then write the minimum, plus a small test if it has logic. Author's benchmark in Claude Code (39 tasks, 5 runs each): • 53% less code • 45% fewer tokens • 41% less time • 26% lower cost • 98% of risky logic ships with a test (68% without) It never cuts validation, security, error handling or accessibility. Install in Claude Code: /plugin marketplace add DietrichGebert/ponytail /plugin install ponytail@ponytail Install in Codex: codex plugin marketplace add DietrichGebert/ponytail codex plugin add ponytail@ponytail Modes: /ponytail lite | full | ultra Subscribe so you don't miss Day 9. Instagram: @princegupta.ai #shorts #claudecode #aiagents #codex #opensource

I pointed my Claude skill at 3 famous repos, and it animated their READMEs 🎨 --- title: I pointed my Claude skill at 3 famous repos, and it animated their READMEs 🎨 published: true description: living-canvas turns any GitHub repo into an animated README banner. Here's what it made for Ponytail, Ollama and uv. tags: showdev, opensource, ai, github --- A README has about two seconds to tell a stranger what a project does. Most banners spend those two seconds on a logo and a gradient. 🫠 I built living-canvas to do better: it's a Claude skill that reads a repo's README, finds the one idea at its heart, and turns it into a short animated story, left to right: the problem → what the project does → the result. To see if it holds up, I pointed it at three projects I love. 👇 🐴 Ponytail: the lazy senior dev Ponytail by Dietrich Gebert is a skill that makes your AI coding agent think like the laziest senior dev in the room, so it writes less code. The story comes straight from Ponytail's README. You ask for a date picker. Without Ponytail, the agent builds a whole calendar by hand: 335 lines. Ponytail first climbs its ladder of checks (does this need to exist? is it already in the codebase? the standard library? a native platform feature?) and finds that every browser already has a date picker. Result: 10 lines. 🪶 🦙 Ollama: open models on your machine Ollama lets you run open models locally, with complete privacy. On the left, every prompt (and your data) goes to someone else's cloud, with an API key and a meter running. In the middle, one command: ollama run gemma4. On the right, the model answers on your own machine at localhost:11434. No key, no cloud, and it works offline. 🔒 !Ollama ⚡ uv: one tool, 10–100x faster uv by Astral is an extremely fast Python package and project manager, written in Rust. Six tools for one Python project (pip, pip-tools, pipx, poetry, pyenv, virtualenv) become one binary. uv add ruff resolves and installs in a blink, and the speed bars show uv's 10–100x lead over pip. 🏎️ !uv 🧩 How it works Every banner follows the same rules: - 🖼️ One self-contained SVG, around 8 KB each for these three - 🚫 No JavaScript, no external fonts or images, so it renders right inside a GitHub README - 🏷️ Every stage is labeled in plain words, so a single screenshot tells the whole story - 📖 Real numbers from the README, never invented ones - ♿ Respects reduced motion: viewers who turn animation off see all three stages at once 🚀 Try it on your repo In Claude Code: git clone shouvik12/living-canvas ~/.claude/skills/living-canvas Then ask: /living-canvas you/your-repo Commit the SVG it gives you and add one line to the top of your README: <p align="center"><img src="assets/your-banner.svg" alt="Your project" width="100%"></p> 🔗 Links - 🎨 living-canvas: shouvik12/living-canvas - 📺 DEV·TV, where living-canvas is currently running in the commercial breaks: shouvik12.github.io - 🐴 Ponytail · 🦙 Ollama · ⚡ uv If you maintain one of these and want the SVG for your README, it's yours. Just say the word. 🙌

I noticed my usage spiking a lot last week but I ultimately thought I was just being negligent, but now it seems it went back to normal. Also which effort level and model are you using? It might be worth taking a look at the skills you have, I noticed improvements after installing the pony tail skill - dietrichgebert/ponytail