Building for the world, together. 🌍 So inspiring to see what our community created at the Google I/O Hackathon hosted by Cerebral Valley and Google DeepMind. Incredible energy, brilliant builders, and innovative projects powered by Gemini 3.5 Flash. Until next time! Check out the winners gallery page: https://goo.gle/4v0Zynj
Google for Developers
Technology, Information and Internet
Mountain View, CA 4,104,589 followers
Join a community of creative developers and learn how to use the latest in technology—from AI and cloud, to mobile & web
About us
Discover the latest technologies, resources, events, and announcements to help you build smarter and ship faster. Explore more at developers.google.com
- Website
-
http://developers.google.com
External link for Google for Developers
- Industry
- Technology, Information and Internet
- Company size
- 10,001+ employees
- Headquarters
- Mountain View, CA
- Specialties
- coding, engineering, firebase, android, cloud, web development, and mobile development
Updates
-
Most AI agent demos reset after one conversation. But what if your agent needs to run for days — pausing, resuming, waiting for human input — without losing context or burning compute? In this episode of The Agent Factory, Smitha Kolan sits down with Addy Osmani, Director at Google Cloud AI, to show what production-grade agent architecture actually looks like using ADK 2.0 and Gemini 3.5 Flash. Three live demos, zero hand-waving: → New hire onboarding coordinator: An event-driven agent that orchestrates IT provisioning, pauses for human actions, and resumes without token bloat. → "AEOS" browser operating system: A coding agent that builds a full window manager, text editor, file browser — and a working version of Doom inside the browser. → 3D asset optimization pipeline: A multi-step agent that takes a 156MB Blender file and compresses it into a fast, browser-ready retro video store scene using Python and GLTF optimization. They also dig into the human side — drawing from Addy's book "Beyond Vibe Coding" to tackle cognitive debt, cognitive surrender, and how dev teams can stay sharp while building with agents. If you're building agents that need to survive beyond a single prompt, this is the episode. Check out this episode and more in the Agent Factory series → https://goo.gle/3QdbgMw
-
Pitch-perfect code meets the beautiful game. Who’s in your starting lineup? ⚽️ SQL: The Goalkeeper 🧤 The ultimate source of truth who keeps a "clean sheet." Organizes the backline, locks down the box, and retrieves the exact data you need the second the team calls for it. C++: The Center Back 🧱 A high-performance powerhouse. Requires elite precision to manually mark every threat, ensuring the foundation remains completely unbreakable under extreme pressure. Python: The Creative Playmaker 🧠 The classic "No. 10." Relies on a massive library of brilliant moves to orchestrate the attack, effortlessly pulling the strings without needing to sprint for every loose ball. Go: The Box-to-Box Midfielder 🏃💨 The engine room. Built for pure stamina, handling thousands of concurrent sprints across the pitch to keep the team’s infrastructure moving in perfect parallel. Kotlin: The Modern Wing-Back ⚡ A polished professional who rarely loses possession thanks to built-in null-safety. Effortlessly shuts down errors and delivers world-class crosses with a clinical, modern touch.
-
Stop fighting with backend infrastructure just to get your AI agents to scale. With Managed Agents in the Gemini API, you can now deploy production-ready agents inside remoteGoogle-hosted sandboxes—with a single API call. Powered by Gemini 3.5 Flash and the Antigravity agent harness, these autonomous agents can plan, reason, write code, and search the web. 🛠️ What you can do right now: • Zero-config environments: Spin up an ephemeral Linux sandbox instantly • State preservation: Pass an environment_id to continuously modify and access files across multi-turn conversations without losing state. • File-based control: Define instructions and specific capabilities using plain markdown files (AGENTS .md and SKILL .md). • Flexible seeding: Mount custom data directly from inline text, Cloud Storage, or public GitHub repositories. The hardest part of building agents is now the part you don't have to build. 📺 Check out the quick walkthrough https://goo.gle/4uab5Q0 and explore the new Agents templates in ai.studio/managed-agents
-
Imagine livedubbing for your audio or video content. With the Gemini API, now you can. Watch Gemini 3.5 Live Translate process speech-to-speech in near real-time, streaming translated audio directly from a video source - going from English to Hindi in seconds! 👇 Watch the full video to see it in action: https://goo.gle/4oeB47z 📝 Read the full blog announcement: https://goo.gle/4fuRlmG
-
Google for Developers reposted this
Lighthouse is getting a new agentic browsing category. Instead of just auditing for human users, it tracks WebMCP schemas and layout shifts to see if AI agents can actually navigate your site. If machine-readability is on your roadmap, this is worth a look → https://goo.gle/3ReK3JD #GoogleIO
-
-
Introducing DiffusionGemma, an experimental open 26B Mixture of Experts model that moves beyond traditional sequential generation to process and generate entire blocks of text simultaneously. DiffusionGemma unlocks new value for developers: 🏃♀️ Generates 1,000+ tokens/sec on an NVIDIA H100 and 700+ tokens/sec on an RTX 5090 🔄 Optimizes non-linear workflows like code infilling, inline editing, and real-time self-correction 📦 Comfortably within 18GB VRAM limits of high-end dedicated consumer GPUs when quantized. 🔧 Supports native integration for MLX, vLLM, Hugging Face, and Unsloth with advanced NVIDIA NVFP4 kernel optimization 📥 Download the Apache 2.0 weights on Hugging Face: https://goo.gle/4uCwBgQ 📖 Explore the developer guide on the blog: https://goo.gle/3RX6zHn
-
-
Your code shouldn't just compile, it should communicate 🗣️ Ensure your identifier names stay clean with these practices: 📏 Use brief names for local scopes 🛑 Drop redundant context 💻 Rely on modern editor type detection Paraphrase your naming conventions to find an easier way to solve readability problems.