Google isn’t trying to win the AI race. They’re trying to own the entire AI Agent ecosystem. While everyone argues ChatGPT vs Claude, Google quietly built: Models → Gemini Pro, Flash, Deep Think, Gemma Design → Stitch, Whisk, Imagen Research → NotebookLM, AI Mode Video → Veo, Flow, Google Vids Coding → Antigravity IDE, Gemini CLI, Jules Agents → A2A, ADK, FileSearch API The scary part? All of these tools talk to each other. That means: 10x faster prototypes End-to-end AI workflows Production-ready agents on GCP The next AI war won’t be model vs model. It’ll be ecosystem vs ecosystem. I mapped this stack out here: gamma.app/?utm_campaign=prom… Save. Share. Build.

8/ 19 C2 commands in total. Highlights: - shell: arbitrary cmds, opt. sudo with the stolen password - load: runs Mach-O / DMG / shell payloads from URL - upload_notes: exfil Apple Notes - filesearch: by ext/size/depth - inject / dyld / load_dylib

Need to search an entire folder tree for text without writing messy recursive code yourself? This utility adds maxDepth control, clean separation of search and traversal, and method overloading so the simple one-folder version still works. Drop it in a module and search up to any depth — or everything with -1. blog.xojo.com/2026/05/04/bui… #Xojo #FileSearch #Recursion #CodeUtility

Episode 6 of Ship AI with Laravel is live! Last episode we built semantic search from scratch. This time we let the AI provider host the vector store and we just plug into it with the FileSearch tool. Upload PDFs, full policy docs, anything you want. The agent searches them and we don't write a single line of search logic. Come build with me 👇

Google isn’t trying to win the AI race. They’re trying to own the entire AI Agent ecosystem. While everyone argues ChatGPT vs Claude, Google quietly built: Models → Gemini Pro, Flash, Deep Think, Gemma Design → Stitch, Whisk, Imagen Research → NotebookLM, AI Mode Video → Veo, Flow, Google Vids Coding → Antigravity IDE, Gemini CLI, Jules Agents → A2A, ADK, FileSearch API The scary part? All of these tools talk to each other. That means: 10x faster prototypes End-to-end AI workflows Production-ready agents on GCP The next AI war won’t be model vs model. It’ll be ecosystem vs ecosystem. Save. Share. Build.

Você pode criar um bot corporativo que recorre a documentos. Você não precisa criar o próprio RAG, pode terceirizar isso pro Google. Ocorre que aparentemente o filesearch faz apenas busca semânica. O ratio, particularmente, possui busca lexical semântica (quando uma súmula X é citada, o match do número é muito importante. Acho que pode haver prejuízo se deixar dependente apenas da busca semântica. Por exemplo: "roubo" e "furto" podem ter embeddings muito próximos no espaço latente por dividirem o mesmo contexto criminal. Juridicamente, são tipificações distintas. A busca lexical atua como um filtro rígido para garantir o termo exato.) Acho que seria útil pra escritórios grandes. Talvez importando toda a documentação possibilite o adv a encontrar um documento específico mais rápido, como por exemplo "Qual foi nossa tese de prescrição intercorrente naquele caso de execução fiscal?" O modelo entende o conceito, responde, e o rounding_metadata vai gerar a citação com o link exato para o arquivo no servidor de onde ele tirou a resposta. Acho que seria útil também subir um edital de licitação complexo e fazer Q&A direto com o texto...

Google isn’t trying to win the AI race. They’re trying to own the entire AI Agent ecosystem. While everyone argues ChatGPT vs Claude, Google quietly built: Models → Gemini Pro, Flash, Deep Think, Gemma Design → Stitch, Whisk, Imagen Research → NotebookLM, AI Mode Video → Veo, Flow, Google Vids Coding → Antigravity IDE, Gemini CLI, Jules Agents → A2A, ADK, FileSearch API The scary part? All of these tools talk to each other. That means: 10x faster prototypes End-to-end AI workflows Production-ready agents on GCP The next AI war won’t be model vs model. It’ll be ecosystem vs ecosystem. I mapped this stack out here: gamma.app/?utm_campaign=prom… Save. Share. Build.

Oookeyブログを更新しました！ 今回はOpenAIのFileSearchの仕様や実装方法を紹介します！ 【OpenAI API】FileSearchについて oookey.co.jp/blog/025_progra…

People of Pi "pi install npm:pi-fff". Just created a package for searching files with @ using the goat @neogoose_btw fff filesearch package. Now you can enjoy typo safe file searching in peace

Since adding lsp support to the code reviewer thought It would make sense too add full file view support. Also messing around with @neogoose_btw fff filesearch really fast and performant. Will take the week to test all new features and will do release prob next week.

Google 不是来赢 AI 比赛的。他们是要掌控整个 AI Agent 生态圈。当大家都在吵 ChatGPT 对 Claude 时，Google 悄悄布局了： 模型 → Gemini Pro、Flash、Deep Think、Gemma 设计 → Stitch、Whisk、Imagen 研究 → NotebookLM、AI Mode 视频 → Veo、Flow、Google Vids 编程 → Antigravity IDE、Gemini CLI、Jules Agent → A2A、ADK、FileSearch API 可怕的是？这些工具能互相对话。意味着： 10 倍速原型开发 端到端 AI 工作流 GCP 上可用的生产级 Agent 下一场 AI 战不是模型对模型， 而是生态对生态。

Google isn’t trying to win the AI race. They’re trying to own the entire AI Agent ecosystem. While everyone argues ChatGPT vs Claude, Google quietly built: Models → Gemini Pro, Flash, Deep Think, Gemma Design → Stitch, Whisk, Imagen Research → NotebookLM, AI Mode Video → Veo, Flow, Google Vids Coding → Antigravity IDE, Gemini CLI, Jules Agents → A2A, ADK, FileSearch API The scary part? All of these tools talk to each other. That means: 10x faster prototypes End-to-end AI workflows Production-ready agents on GCP The next AI war won’t be model vs model. It’ll be ecosystem vs ecosystem. Save. Share. Build.

🚀 منظومة جوجل المتكاملة للذكاء الاصطناعي : 🧠 النماذج : 1ـ Gemini 3 Pro: نموذج الاستنتاج الأكثر تقدماً من جوجل. 2ـ Gemini 3 Flash: ذكاء رائد مصمم للسرعة والكفاءة. 3ـ Gemini 3.1 Flash Lite: النموذج الأسرع والأكثر توفيراً للتكلفة في عائلة Gemini. 4ـ Gemini (Deep Thinking): نموذج تفكير عميق مخصص للأبحاث المعززة بالوكلاء الذكيين. 5ـ Gemma: نماذج مفتوحة المصدر وخفيفة الوزن من أبحاث DeepMind. ـــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــ 🔍 البحث : 6ـ NotebookLM: مساعد بحث مدعوم بالذكاء الاصطناعي لتنظيم الأفكار. 7ـ Pomelli: لتوليد أفكار الحملات الاجتماعية للعلامات التجارية. 8ـ AI Mode: استدلال متقدم وتعدد وسائط لعمليات بحث قوية. ـــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــ 🎨 التصميم : 9ـ Stitch: تحويل النصوص إلى تصميمات واجهة مستخدم (UI) معقدة. 10ـ Whisk: استخدام الصور كأوامر نصية لتصور الأفكار وتجسيدها. 11ـ Nano Banana: توليد الصور باستخدام قدرات الاستنتاج والمعرفة لدى Gemini. ـــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــ 🎬 الفيديو : 12ـ Veo 3.1: أعلى جودة لتوليد الفيديو من النصوص. 13ـ Flow: أداة سرد قصصي لإنشاء مقاطع ومشاهد سينمائية. 14ـ Google Vids: إنشاء فيديوهات مدعومة بالذكاء الاصطناعي مخصصة للعمل. ـــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــ 💻 البرمجة : 15ـ Gemini CLI: وكيل مفتوح المصدر يجلب نماذج Gemini إلى واجهة الأوامر (Terminal) الخاصة بك. 16ـ Antigravity: بيئة تطوير متكاملة (IDE) مدعومة بالذكاء الاصطناعي مع وكلاء مستقلين. 17ـ Jules: مساعد برمجة غير متزامن للمستودعات البرمجية. ـــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــــ 🤖 الوكلاء الذكيون (AI Agents) 18ـ Google ADK: إطار عمل لتطوير وكلاء ذكاء اصطناعي قابلين للتوسع. 19ـ Google A2A: بروتوكول اتصال بين الوكلاء المتعددين بغض النظر عن أطر العمل المستخدمة. 20ـ FileSearch API: استخدام خط معالجة RAG كامل وجاهز للإعداد عبر واجهة برمجة تطبيقات بسيطة.

It was in 2025 when I heard about Retrieval-Augmented Generation (RAG), at a time when I used to have conversations with the likes of @ashioyajotham_ where in one of his articles, he suggested that we should not be saying that AI generates hallucinations but rather call them confabulations; AI generates output based on what it was trained on and will try to 'fill in the gaps' if it knows little about the topic at hand. I then went on to learn about RAG and tried to implement the process in one of my projects. Fast forward to 2026, you no longer need to know the nitty-gritty parts of RAG. All you need to know is about Google's File Search Tool and how it works. Take a read on what it's all about and how I built a project using this tool 👇🏿 Built using @angular, #Genkit, @Firebase and Gemini API's #FileSearch tool.

Google isn’t trying to win the AI race. They’re trying to own the entire AI Agent ecosystem. While everyone argues ChatGPT vs Claude, Google quietly built: Models → Gemini Pro, Flash, Deep Think, Gemma Design → Stitch, Whisk, Imagen Research → NotebookLM, AI Mode Video → Veo, Flow, Google Vids Coding → Antigravity IDE, Gemini CLI, Jules Agents → A2A, ADK, FileSearch API The scary part? All of these tools talk to each other. That means: 10x faster prototypes End-to-end AI workflows Production-ready agents on GCP The next AI war won’t be model vs model. It’ll be ecosystem vs ecosystem. Save. Share. Build.

Everyone’s debating ChatGPT vs Claude Google’s Full Stack AI Ecosystem From Research to Development Google is quietly building one of the most complete AI ecosystems Instead of just releasing models, they are creating an entire stack that covers models, coding, agents, research, design & video generation 🔹 Models: Gemini 3 Flash, Gemini 3 Pro, Gemini 3 Lite, Gemini 2.5 Deep 🔹 Coding: Gemini CLI, Jules, Antigravity 🔹 AI Agents: Google ADK, Google A2A, FileSearch API 🔹 Research: NotebookLM, AI Mode, Pomelli 🔹 Design: Stitch, Whisk, Nanobanana 🔹 Video: Veo 3.1, Flow, Google Vids The interesting shift is that Google is not just focusing on AI models, but on building tools, agents & developer frameworks that make AI easier to integrate into real products We are moving from “AI tools” → “AI ecosystems” The real advantage in the AI race may not just be the best model but the best ecosystem around it What part of this stack excites you the most?