Ξlpida

Google's Elephant Goldfish Model is a plugin that reviews software designs without sharing the conversation that produced them. One agent keeps the project context and writes the draft. Separate reviewers receive the draft and its referenced sources. They look for missing assumptions and unclear requirements. You decide which findings to address and approve the design before implementation.

Before your AI builds the app, give its plan to reviewers who weren't in the conversation. The Elephant-Goldfish Model: a free open-source plugin for Claude Code and Codex. Comment PLAN and I'll send it over to you.

🚨 غوغل تطلق مجاناً السلاح السري الذي يستخدمه آلاف مهندسيها يومياً لمنع الذكاء الاصطناعي من كتابة أسرع كود مليء بالأخطاء! 🧵 ​الأداة تُدعى "Elephant-Goldfish Model"، وفكرتها بسيطة لكنها عبقرية بشكل لا يُصدق: ​❌ لا كود يُكتب حتى تكتمل الخطة ويصمد التصميم! ❌ لا يُسمح للذكاء الاصطناعي بتقييم عمله بنفسه! ❌ لا ضياع للسياق في منتصف المشروع! ​إليك كيف تُدار اللعبة داخل غوغل 👇 ​🧠 الفيل (Elephant): جلسة ذكاء اصطناعي واحدة طويلة المدى، تتذكر كل شيء: المعمارية، الهدف، والقرارات الصغيرة. دورها: تحاورك، تناقشك، وتُوثّق كل تفصيلة.. لكنها لا تكتب سطراً واحداً من الكود! ​🐟 السمكة الذهبية (Goldfish): جلسات ذكاء اصطناعي جديدة كلياً وبلا أي ذاكرة، لا تعرف شيئاً عن نقاشاتك السابقة حتى لا ترث أخطائك الذهنية أو تحيزك. ​🔄 كيف تعمل هذه المنظومة العبقرية؟ ​1️⃣ العصف الذهني: مناقشة بينك وبين "الفيل" بحظر تام لكتابة الكود. 2️⃣ المسودة: صياغة وثيقة تصميم دقيقة تسبق التنفيذ. 3️⃣ اختبار الفهم : تطلب من "سمكة ذهبية" إعادة شرح التصميم، إذا أخطأت، يُعتبر المستند غامضاً ويُعاد صياغته! 4️⃣ الهجوم المضاد: عدة أسماك ذهبية تهاجم التصميم كمراجعين ماليين للبحث عن الثغرات والحالات الخاصة. 5️⃣ القرار النهائي: سمكة ذهبية أخيرة تعطي الضوء الأخضر بناءً على وضوح مسارات الملفات والـ Signatures. ​🔥 لماذا يُعد هذا الأمر تحولاً جذرياً؟ ​يقول فريق غوغل: "الذكاء الاصطناعي جعل كتابة الكود أسرع، لكنه عجّل أيضاً بنشر الأكواد العشوائية والأخطاء. السرعة بلا حواجز حماية مجرد مضاعفة لأخطائك". ​هذا النموذج يعكس الآية: يتم تدمير واختبار التصميم واكتشاف عيوبه قبل كتابة سطر كود واحد! ​💡 الخبر الأفضل: الأداة مفتوحة المصدر مجاناً (ترخيص Apache 2.0)، وتتضمن كافة القواعد وشخصيات المراجعين. (ملاحظة: تعمل حالياً داخل Antigravity مع دعم باقي البيئات قريباً). ​📌 احفظ التغريدة للرجوع إليها 🔖 رابط المستودع على GitHub تجدونه أسفل التغريدة 👇

Google just open-sourced the method thousands of its engineers use every day to stop AI doing sloppy work. It's called the Elephant-Goldfish Model. Google built it for code, and it works for anything you ask an AI to plan or write. The chat that plans never builds. A second chat that hasn't seen the conversation checks the plan. How to run it in ChatGPT, Claude or Gemini: 1) The elephant. Open a chat and talk through the thing you want: the trip, the business, the kitchen, the essay. Tell it: "No output yet. Ask me questions until you could write the brief yourself." Give it 20 minutes. 2) The brief. "Write a one-page brief: what I want, why, what's already decided, what's still open, and the steps in order." Read it, fix what's wrong, copy it. 3) The goldfish. Open a new chat with no history (Temporary Chat in ChatGPT). Paste the brief and nothing else, then run three checks: "Summarise this back to me in five lines, as if you're the one doing it." If it gets it wrong, fix the brief. "You're the most sceptical person in the room. List every assumption here that isn't proven, and the three places this goes wrong." "Could someone start step one right now with only this page? If not, what's missing?" 4) Fix the brief with what the goldfish found. Paste it into a chat and say: build it. The first chat has been talking to you for 20 minutes and fills every gap with what it thinks you meant. The second chat has only the page in front of it, so it can only find gaps that are on the page. Google's line on the goal: the AI produces what you wanted the first time you press enter. The plugin is free on GitHub, search elephant-goldfish-model.

Elephant-Goldfish Model Thousands of Google engineers use Elephant-Goldfish Model to design and review software before writing code. The open source plugin separates design drafting from independent review using AI agents. Elephant drafts specs with full project context. Fresh Goldfish subagents review the draft and references without prior chat history, exposing missing assumptions and implementation risks. It provides skills for brainstorming, design reviews and detecting drift between docs and code. It supports Antigravity, Claude Code and Codex. github.com/google/elephant-g…

Google just open-sourced the plugin thousands of its engineers use every day to stop AI from shipping sloppy code. It's called the Elephant-Goldfish Model. No code until the plan survives. No AI grading its own homework. No context collapse halfway through a feature. The idea is simple and kind of brilliant. The Elephant is one long-running AI session that remembers everything: the architecture, the intent, every small decision. It writes zero code. It interviews you, argues with you, and keeps the record. The Goldfish are fresh AI sessions with no memory at all. They never saw your design conversation, so they don't inherit its blind spots. Here's how it works: → egm-brainstorm: you and the Elephant debate options under a strict no-code rule → egm-draft: you produce a Markdown design doc with the problem, the plan, the alternatives, and the exact files to change → egm-review: one Goldfish has to explain the design back. If it gets it wrong, the doc is ambiguous and goes back for redrafting → Then several Goldfish attack it as hostile reviewers hunting edge cases, hidden dependencies and logic gaps → A final Goldfish makes the go/no-go call: are file paths, data structures and function signatures all present and unambiguous? Here's why this matters: Google's team says AI made coding faster but also sped up slop, bugs and session context collapse. Speed without guardrails just scales your mistakes. This flips the order. The design gets attacked before a single line exists. Here's the wildest part: This isn't a research demo. Google says thousands of Googlers run it daily to ship production code. The repo ships the Markdown rules, the reviewer personas and the review workflows. Free, Apache 2.0. The honest trade-off: it only runs inside Antigravity for now, with other environments coming. And security isn't on the default reviewer list. You'd add your own security Goldfish. github.com/google/elephant-g…

Google's Elephant-Goldfish Model is a free, open-source plugin for Claude Code, Codex and Antigravity. Google says thousands of its engineers use it. It's for anyone whose agent starts coding before the plan holds up. What I'd copy into my own setup: 1. One agent with your full project context drafts the spec before any code gets written. 2. Fresh subagents that never saw the chat review it. They catch the assumptions you never said out loud. 3. You make the trade-offs and approve the design. Only then does the agent build. 4. Skip it for small, clear bug fixes. Use it when a feature needs real design calls. 5. Later, run the drift report to check your docs still match the code. Repo: google/elephant-goldfish-model on GitHub. Worth the read. nitter.gq/VaibhavSisinty/status/…

GOOGLE ENGINEERS REVIEW THEIR CODE WITH A F*CKING GOLDFISH... it's called the elephant-goldfish model. open source, 3.2K stars → the elephant: your main agent. knows the whole project, writes the design doc → the goldfish: fresh subagents with zero memory of your chat. they read the doc cold and catch everything the elephant "just knew" → you: pick the trade-offs, approve the spec, then it builds skills for brainstorm, draft, review, resolve, plus a drift report that checks your docs against the code works in claude code, codex and antigravity does your agent review its own work?