Gemini is now even smarter – New tools announced at Google I/O 2025

As Google I/O 2025 gets underway, today’s keynote has unveiled a series of significant upgrades for Gemini. Eight distinct enhancements have been introduced, signaling a focused evolution of Google’s AI model. These advancements touch upon areas such as Gemini’s multimodal understanding, its integration across more platforms, and improvements in its reasoning capabilities. Further developments include expanded developer tools for Gemini, refined efficiency for broader deployment, and new ways for users to interact with the AI. This overview will detail these specific upgrades and their potential impact on how Gemini functions and is utilized.

Google Beam

Google Beam is an AI-first 3D video platform. This technology integrates Google’s AI video model with a light field display to create a virtual conversation experience designed for a more natural feel. Users can perceive the dimensionality and depth of the remote participant without requiring a headset or specialized eyewear.

The first Google Beam devices are slated for release early this year, in collaboration with HP. This development could potentially alter workplace communications by introducing an added layer of depth to virtual interactions.

Agentic Shopping

This new feature, called Agentic shopping, introduces a “buy for me” capability within AI Mode. Users will be able to track the price of any product and set a desired spending limit. Once the “buy for me” option is activated, the system can automatically place the order via Google Pay when the item reaches the specified price. This functionality is designed to assist users in securing purchases at their preferred price point.

Inbox cleanup

This is an automated process that helps you manage your email by identifying and removing unwanted or unnecessary messages based on specific criteria you set. It’s about decluttering your digital inbox to make it more organized and functional.The new Inbox Cleanup feature lets you simply ask Gemini to clean your inbox using a conversational prompt. Google gave the example, “Delete all of my unread emails from The Groomed Paw from last year,” and Gemini would take care of it in seconds.

Free Gemini Live Multimodal Access

A significant advancement in AI voice assistants lies in their multimodality, which allows them to understand context beyond just spoken words. This means the assistant, already capable of conversational interaction, gains an additional layer of awareness from your surroundings. Google is now making Gemini Live, with its camera and screen sharing capabilities, freely accessible to all users on both Android and iOS platforms. This move aims to broaden the reach of this contextual assistance.

AI Mode for All

A major focus at Google I/O revolved around AI Mode, Google’s experimental search interface that blends traditional search with conversational chatbot capabilities. Several promising updates were highlighted, notably the integration of Project Mariner. This research prototype, built on Gemini 2.0, is designed to automate various tasks directly within your browser. This includes the ability to handle ticket purchases and make reservations on your behalf, significantly streamlining online interactions.

Veo 3 and Flow

In the domain of AI video generation, Veo 3 presents a notable development. This tool differentiates itself by incorporating audio capabilities, allowing for the generation of background noises and even dialogue within video clips. This functionality enhances the potential utility of generated clips for various content creation needs.

Alongside Veo 3, Google also introduced Flow, an AI filmmaking tool. Flow combines Veo 3 with the recently released Imagen 4 to facilitate the creation of high-quality videos. While these specific features may not see daily use by a broad audience, their introduction marks a step forward in advanced AI-driven content creation tools within the industry.

Jules: Asynchronous Coding Agent

Google has moved its Jules asynchronous coding agent from a Labs prototype into public beta, making it broadly available without a waitlist. For those involved in coding, Jules integrates with existing repositories, including GitHub.

Operating autonomously, Jules is designed to reason through decisions to perform tasks such as debugging code, building new features, and writing tests. It also offers the ability to preview its proposed changes before they are committed. Regarding privacy, Google states that Jules does not train on private codebases, instead cloning the codebase into a secure Google Cloud virtual machine for its operations. This release positions Jules in a similar space to OpenAI’s Codex, which was announced recently.

Android XR

Perhaps a counter option to Meta Rayban glasses, Android XR outlines a path for integrating AI into the physical world for enhanced assistance. The associated glasses feature microphones, a camera, and speakers, allowing them to perceive your environment. Paired with Gemini, this provides the AI with significant context for daily activities, enabling more robust support. An optional in-lens display can also show private information.