Developer Keynote (Google I/O '26)

Advancements in AI Models and Antigravity Platform 00:15

“We've got the new Omni model, the new Gemini 3.5 series models, and last month, we also shipped Gemma 4 under an Apache 2 license.”

• Josh Woodward introduces significant advancements in AI models, highlighting the release of the Omni model, the Gemini 3.5 series, and the notable Gemma 4, which operates under an Apache 2 license.

• Gemma 4 has received an overwhelming response, achieving 100 million downloads within the first month and surpassing half a billion downloads overall.

• This model is designed for advanced reasoning and agentic workflows, making it versatile enough to run offline on mobile devices and applicable in various fields including robotics and space exploration.

Transitioning to Agentic Tools 01:15

“The big shift is our move towards agents, from AI that simply assists you to agents that help you get things done under your direction.”

• A key focus is the transition from traditional AI assistance to agentic AI, which actively helps users complete tasks more efficiently.

• Central to this initiative is Google Antigravity, a platform designed for developing and integrating these intelligent agents into applications.

• Developers are encouraged to build applications utilizing this platform, whether they choose Google's infrastructure or their own, and can do so on both Android and web platforms.

Introducing Managed Agents in Development 02:21

“We’re putting the full power of agents directly into your hands no matter what you’re building.”

• Logan Kilpatrick emphasizes the availability of managed agents, which empower developers to enhance their applications through complex tasks such as data science or research.

• The Interactions API, introduced last December, provides a unified interface for models and agents and has already been employed internally within Google for sophisticated projects.

• The Antigravity Harness is now made available to developers, allowing easy integration of agents into projects via the Gemini API while ensuring that workloads remain secure and manageable.

Real-World Application of Managed Agents 04:09

“Managed agents are available starting today, and thanks to our ecosystem partners, you can start building with your preferred stack from day one.”

• Illustrating the power of managed agents, a case example is provided featuring the Stitch team, which successfully integrated managed agents to enhance user experience by connecting to GitHub and generating tailored design resources.

• The ease of scaling with managed agents means that applications can efficiently handle increasing user demands while streamlining operational tasks for developers.

Building and Customizing Agents in AI Studio 05:21

“These open-source agent templates come preloaded with instructions, skills, and tools as markdown so you can customize the source files.”

• Developers can rapidly prototype agents with ready-to-use templates in the AI Studio playground, allowing for easy customization and immediate deployment.

• An example is given with an AI talk radio agent, demonstrating how a developer can generate an entire radio show autonomously by simply providing a topic.

• The standout feature is the use of markdown for defining agent functionalities, simplifying the interaction and orchestration of complex workflows without needing extensive programming.

Deploying Applications in AI Studio 09:45

“Deploying from AI Studio to Cloud Run is just a few clicks.”

• The presentation outlines how straightforward it is for developers to deploy applications created in AI Studio, with minimal steps involved, including automatic project generation for new users.

• The session highlights new integrations supporting full-stack app development, including Google Workspace, allowing for seamless operations between various Google applications while enhancing user interaction through agents.

• By enabling the development of Android apps directly within AI Studio, developers can rapidly transition their ideas into functional applications without extensive setup requirements.

Android App Development with AI Studio 11:57

"You can preview your new app with the Android Emulator directly in AI Studio."

• The presenter demonstrates how to develop an Android app using AI Studio, built entirely in Kotlin, and emphasizes the streamlined process from idea to implementation.

• Users can utilize the Android Emulator within AI Studio to view their creations, and when ready, they can install the app on a personal device simply by connecting their phone.

• AI Studio now supports Google Play Store publishing, allowing developers to push their app to the test track within minutes after connecting their Play Developer Account.

Mobile Experience and Antigravity SDK 13:10

"We're bringing the AI Studio build experience to your pocket with our brand-new mobile app."

• Google introduces a new mobile app that extends the capabilities of AI Studio, enabling developers to build apps from anywhere without being tied to their desks.

• For those who seek greater customization and control, the Antigravity SDK is launched, offering the same agent harness optimized for Gemini with enhanced flexibility for deployment.

• The lifecycle of an app is discussed, highlighting that building a successful agent requires ongoing development beyond initial creation.

One-Click Export to Antigravity 14:21

"We are porting the complete file system and every single ounce of context."

• The speaker announces a one-click export feature that allows developers to transfer their projects from AI Studio to Antigravity without losing context, which significantly simplifies the development process.

• This new capability means developers can easily continue their work without the hassle traditionally associated with transferring projects between platforms.

Introduction of Antigravity 2.0 14:52

"We're making it easier to build with agents."

• Anshul Ramachandran introduces Antigravity 2.0, a desktop application that prioritizes agent-driven development and allows multiple agents to operate on various projects concurrently, enhancing collaboration and efficiency.

• The introduction of Dynamic Subagents allows agents to create specialized helpers that can tackle specific tasks, improving overall productivity and task management.

Scheduled Tasks Feature 16:18

"You can now use standard cron scheduling to put your agents on Autopilot."

• Antigravity 2.0 includes a Scheduled Tasks feature that automates recurring tasks, allowing agents to operate autonomously on predefined schedules, such as summarizing pull requests or monitoring cloud health.

• This feature aims to make agents more proactive in managing tasks without continuous human intervention.

Antigravity CLI Announcement 19:31

"It's a lightweight way to spin up the same Antigravity agents right from the terminal."

• The Antigravity Command Line Interface (CLI) is unveiled, providing developers with the flexibility to manage Antigravity agents from the terminal, catering to those who prefer command-line interactions.

• The CLI is designed to be adaptable to personal workflows, themes, and key bindings, ensuring a customizable experience for developers working in various environments.

Real-Time Monitoring in Antigravity 21:02

"Let's take a step back for a minute."

• Kevin demonstrates the real-time capabilities of Antigravity by checking the status of a training job through the CLI, highlighting the ease of monitoring progress and ensuring that training is proceeding healthily.

• Antigravity allows immediate access to task logs and outputs, offering insights into ongoing processes and facilitating quick checks on system health.

Execution of Fine-Tuned Models 22:45

"We just fine-tuned Gemma 4 on stage, which is pretty remarkable."

• The presentation showcased the execution of a fine-tuned model by deploying it to a laptop, illustrating the practical application of the model in real-time.

• The speaker initiated a new conversation in Antigravity 2.0 to interact with the deployed fine-tuned model, demonstrating its effectiveness and efficiency in handling commands without unnecessary additional information.

Unifying Development Platforms 23:52

"Starting today, the Antigravity CLI is available to all Gemini CLI users."

• The Antigravity CLI has been unified as the primary platform for agent-first development, streamlining the experiences of users transitioning from the Gemini CLI.

• Migration guides have been published to assist developers in porting their existing custom skills to the new system, ensuring a seamless transition and continued support for user feedback.

Simplified Integration and Specialization 24:50

"We're also making it easier to push agents into highly specialized fields by introducing domain-specific skills bundles."

• Antigravity is designed to be stack agnostic, allowing developers to use their preferred tools without facing vendor lock-in.

• A new science skill bundle has been introduced to enhance agents' capabilities in health, biology, and scientific research, reflecting a commitment to support specialized development areas.

Enhancing Android Development with Antigravity 26:01

"We want to make it fast and easy to build high-quality Kotlin Android apps."

• Official Android support has been added to Antigravity, improving the efficiency of delivering high-performance applications across various devices.

• The introduction of the Android CLI aims to simplify the setup process for developers and facilitate the use of the most recent Android innovations.

Leveraging AI in Travel Applications 27:34

"Using Antigravity, let's show how you can build an augmented experience today for Display Glasses."

• A travel app was demonstrated, showcasing features like viewing travel plans and transcribing audio diary entries that utilize the Gemini Nano 4 on-device model.

• The presentation highlighted how users can access updated flight and travel information directly on smart glasses, enhancing the user experience, especially in high-demand situations like airports.

Utilizing Android CLI Skills 29:18

"By running Android Studio alongside Antigravity, the agent can leverage these capabilities to run tasks much more quickly."

• Internal testing indicated that agents using Android CLI skills achieved significant efficiency improvements, completing tasks much faster and with reduced resource consumption.

• The integration allows the agent to function effectively, utilizing recommended skills and Android Studio's capabilities for analyzing file issues and managing dependencies.

Performance Analysis in Development 33:19

"Let's get that fixed. I'll go to the agent, and I'll say, okay Gemini, fix and analyze my app's optimization and performance."

• The speaker addressed application performance challenges, utilizing the agent to conduct a comprehensive review of optimization and crash reports from production.

• This included preparing for potential memory limits changes anticipated in upcoming Android versions to ensure the application's reliability and efficiency.

R8 Performance Optimization 33:52

"Effective R8 configurations lead to fewer ANRs, reduced app size, and quicker startup times."

• R8 enhances app performance by removing unused code and resources while rewriting bytecode to optimize run-time efficiency.

• Proper configurations of R8 can significantly diminish application not responding (ANR) incidents, shrink the overall app size, and accelerate startup times.

• The agent utilized R8’s analyzer to recommend changes in the build configuration, maximizing R8's capabilities.

Implementation of Deep Links 35:02

"In the latest version of Android Studio, we have a dedicated App Links Assistant to help you implement new deep links for your app."

• Deep links can increase user engagement by enabling direct navigation to specific content in an app, such as upcoming trips in a travel application.

• Android Studio's App Links Assistant simplifies the creation of deep links by offering tools to manage URL mappings and intent actions effectively.

• The agent analyzed the app's code and the URL to suggest a customized routing logic for implementation.

Direct App Updates to Google Play 36:30

"Now, you can publish updates of your app from Android Studio directly to Google Play."

• Developers can upload new app versions effortlessly from Android Studio to Google Play with just a few clicks, streamlining the update process.

• The integration allows developers to manage the internal test track while preparing the Play Store page, optimizing release procedures for testing.

Migration Assistant for Cross-Platform Development 37:12

"The Migration Assistant is coming to Android Studio later this year."

• A forthcoming Migration Assistant in Android Studio aims to simplify the transition and expansion of applications from iOS to Android for a broader audience.

• The assistant offers options for AI-guided decisions or user-led migration, featuring natural-language instructions for validating and executing app migrations.

• Migrating code and assets such as strings and images can be automated, enhancing efficiency significantly for developers.

Powerful Web Development Tools with AI 41:53

"AI agents are transforming the landscape of development everywhere, and nowhere is that transformation happening faster than on the web."

• The introduction of AI-powered tools enables developers to create and innovate web experiences more rapidly and with greater confidence.

• Baseline equips developers with a comprehensive understanding of web platform features across major browsers, addressing the challenge of feature implementation.

• New tools, like Modern Web Guidance, provide AI agents with blueprints for utilizing the latest web platform features, allowing developers to stay in sync with modern technologies.

Integrating Agents with WebMCP 46:00

"With WebMCP, we can now make this webpage agent-ready in minutes."

• The WebMCP (Web Model Communication Protocol) initiative allows developers to enhance their web pages to be more compatible with browser-based agents.

• This framework enables developers to expose web capabilities that agents can utilize, which increases the precision and reliability of interactions between agents and web pages.

• A practical demonstration showcases the ease of using Modern Web Guidance, where a developer simply asks an agent to implement WebMCP tools for a car configurator.

• The Modern Web Guidance system has been designed to be efficient and effective, boasting a 37 percentage point improvement in guided web development tasks compared to unguided coding.

Leveraging Development Tools for Agents 48:50

"With WebMCP, Gemini and Chrome provide agents with more tools to interact with the website."

• The interaction of Gemini with Chrome illustrates how browser agents can streamline complex tasks by generating a plan based on user prompts.

• Enhancements such as the agent's ability to automatically generate car configurations demonstrate the utility of the integration, allowing users to specify detailed requirements for their builds.

• The demonstration illustrates that these tools are now tailored for the agentic web, enabling seamless automation in web development processes.

Enhancements in Developer Experience with Chrome DevTools 51:20

"Now, your agent can finally see how the code it has written performs at runtime."

• New functionalities in Chrome DevTools enable coding agents to observe and assess the performance of the code they generate in real-time, creating a feedback loop that enhances debugging and validation processes.

• The introduction of features like Lighthouse, adjusted for agentic browsing, allows agents to conduct comprehensive health checks on web pages, validating essential components made available through WebMCP tools.

• This capability means that agents can autonomously read reports, suggest solutions, and re-run checks, significantly easing the developer's workload.

Interactive Elements via HTML-in-Canvas 54:35

"The new HTML-in-Canvas API allows you to integrate real DOM elements directly into your canvas environment."

• The novel HTML-in-Canvas API enables the embedding of interactive HTML elements within a canvas, merging accessibility with visual sophistication in web design.

• This breakthrough simplifies the implementation of interactive features, making elements searchable and selectable, thus enhancing user experience across applications.

• By fostering interaction between complex visuals and accessible elements, this technology opens new avenues for creativity in web development, encouraging developers to explore innovative interfaces.

Future of Agentic Web Development 55:58

"In this new era of agentic web development, there’s never been a more inspiring time to create."

• The advancements highlighted in the presentation represent a transformative shift in web development practices, emphasizing the use of tools like Modern Web Guidance and HTML-in-Canvas.

• These tools not only facilitate the quick implementation of features but also expand the possibilities for creating accessible, interactive, and engaging user interfaces.

• Developers are now better equipped to harness these technologies to push the boundaries of creativity and functionality within their web projects.

The Role of AI in the Future of Development 56:26

“The future of the web is imagined by you and supercharged by AI.”

• This statement emphasizes that the creative vision of developers will be enhanced by artificial intelligence, suggesting a collaborative relationship between human ingenuity and technological advancement.

Transforming the Developer Experience 56:50

“Agents completely transform the developer experience.”

• The introduction of intelligent agents is revolutionizing how developers work, allowing them to focus more on innovative ideas while delegating the more labor-intensive tasks to these AI-driven tools.

Launch of the Build with Gemini XPRIZE Hackathon 57:16

“Today, we are officially launching the Build with Gemini XPRIZE Hackathon.”

• This global initiative invites developers to participate in creating applications aimed at solving significant real-world problems, with a total of $2 million in prizes to incentivize innovation.

Aims and Goals of the Hackathon 57:33

“Pick a problem worth solving, build with Gemini, and let's all try to positively impact the lives of 1 billion people.”

• The hackathon encourages participants to focus on meaningful challenges, leveraging the capabilities of Gemini to create solutions that can make a substantial difference on a large scale.

New Google AI Ultra Plan and Promotional Offer 57:47

“...we announced a new $100 per month Google AI Ultra plan.”

• The new pricing plan, which has been adjusted based on user feedback, provides powerful resources for developers. Additionally, a promotional offer of $100 in bonus credits is available for subscribers, enhancing their ability to explore and innovate during the holiday season.

Call to Action for Developers 58:14

“You can visit io.google for the livestream sessions and all the on-demand content we’ll be releasing over the next few days.”

• Developers are encouraged to engage with the available content and resources, ensuring they have access to the latest tools and insights from the conference to further their development endeavors.