What you need to know
- Google is updating the Gemini app with a new interface and Gemini 3 Pro support.
- Google AI Ultra subscribers can try the Gemini Agent, a tool that can complete actions within the Google ecosystem and on the web on your behalf.
- The Gemini app is also gaining two generative interfaces, which are custom views coded in real time by Gemini 3 based on your request.
With the arrival of Gemini 3, Google introduces a sophisticated AI model that enhances multimodal understanding and agentic coding capabilities. This latest iteration is now live within the Gemini app, showcasing a newly designed interface that adapts in real-time to generate custom user interface elements tailored to individual requests.
Among the standout features is the Gemini Agent, an experimental tool that streamlines the completion of multi-step tasks directly within the app. This innovation is a direct evolution of Project Mariner, a research prototype that explored the potential of an AI agent to navigate the web on behalf of users. Insights gleaned from Project Mariner, combined with the intelligence of Gemini 3, have culminated in the creation of the Gemini Agent.
This feature seamlessly integrates with Google Workspace applications and tools such as Deep Research and Canvas, enabling it to handle complex requests with ease. With the added capability of live web browsing, the Gemini Agent represents a significant leap toward realizing a true generalist agent.
Google provides illustrative examples of the Gemini Agent’s potential, including managing calendar events, setting reminders, and organizing Gmail inboxes. While the agent can perform actions autonomously, it prioritizes user control by seeking confirmation before executing significant changes, such as completing purchases or deleting emails.
Additionally, the Gemini app introduces two innovative generative interfaces: visual layouts and dynamic views. Visual layouts offer an immersive, magazine-style presentation, enriched with photos and modular components. In contrast, dynamic views leverage Gemini 3 to construct an interactive user interface in real-time, facilitating exploration and learning from the content surfaced by Gemini. Both interfaces are designed to be customizable through further prompts, enhancing user engagement.
As of today, users can experiment with Gemini 3 Pro by selecting the “Thinking” model, available to both free and paid subscribers. However, the Gemini Agent is currently exclusive to Ultra subscribers in the U.S. within the web app, with Google AI Plus and Pro users enjoying higher rate limits.
