Google I/O '25 Developer Keynote: Gemini, Android, Web, and AI Advancements

Building with Gemini Across Platforms

• 🚀 Gemini 2.5 Pro and Flash are central to new developer tools, enabling rapid prototyping and integration across Android, web, and beyond.
• 💡 Project Stitch is an experimental tool that generates app designs from prompts, allowing for quick iteration and code export.
• ⚡ Google AI Studio now integrates Gemini 2.5 Pro's native code editor, streamlining the creation of web applications and AI-powered games.
• 🛠️ The Google GenAI SDK now natively supports Model Context Protocol (MCP) definitions, simplifying the development of AI apps.

AI Advancements in Android Development

• 📱 Androidify showcases AI capabilities by creating personalized Android bots from selfies using Gemini's multimodal and Imagine 3 models.
• 🧠 On-device AI powered by Gemini Nano offers APIs for summarization, rewriting, and image description, ensuring privacy and performance.
• ✨ Material 3 Expressive and Live Updates enhance app delightfulness and provide timely information to users.
• 💻 Jetpack Compose improvements, including rebuilt navigation libraries and new alpha features, boost developer productivity and UI consistency across devices.
• 🤖 Gemini in Android Studio assists with writing end-to-end tests using natural language and automates dependency updates, streamlining the development lifecycle.

Enhancing the Web with Chrome and AI

• 🌐 New Chrome 135 CSS primitives simplify the creation of complex UI elements like carousels and hover cards, reducing reliance on JavaScript.
• 🔍 Chrome DevTools now feature AI assistance, allowing developers to debug issues using natural language and apply fixes directly within the tool.
• 💡 Gemini Nano and Chrome's built-in AI APIs enable developers to create AI features on-device, ensuring data privacy and cost-effective scaling.
• 🖼️ Multimodal capabilities in Gemini Nano allow for audio and image input, demonstrated by an AI usher that can locate seats from a ticket photo.

Firebase Studio and Gemma for Developers

• 🏗️ Firebase Studio, an AI-powered workspace, allows users to generate fully functional apps from prompts, now integrating with Figma designs via Builder.io.
• 🧩 Gemini in Firebase Studio can generate individual components, hook up routing, and even update placeholder data, breaking down changes into manageable steps.
• 🚀 Gemma 3N, an open model optimized for low RAM devices, offers faster performance and multimodal capabilities, available in Google AI Studio and open-source tools.
• 🏥 Mad Gemma is a collection of open models for multimodal text and image understanding, promising advancements in healthcare applications.
• 🗣️ Sign Gemma models translate sign language to spoken language text, enhancing accessibility, while Dolphin Gemma aids researchers in understanding dolphin communication.

Future of Development and AI Integration

• ☁️ Firebase Studio now automatically provisions backends for apps requiring databases or authentication, deploying directly to Firebase App Hosting.
• 🔧 Developers can fine-tune Gemma models using tools like Unsllo and build UIs with AI-first Colab experiences for personalized applications.
• 🌐 The Gemmaverse showcases tens of thousands of community-created model variants, highlighting Gemma's adaptability and broad language support.
• 🚀 The keynote concludes by emphasizing the power of new AI tools and platforms, encouraging developers to build the next generation of applications.