Back
Technology

Google Gemini Development: Proactive Agents, Voice Changes, and New Features Emerge

Generated on: Last updated: 15 sources
View source

Gemini Assistant: Major Updates on the Horizon

A comprehensive look at upcoming features, including proactive agents, voice changes, and expanded platform integration, based on code analysis and internal documents.

Proactive Assistance & Agent Features

Proactive Assistance

Code analysis of Google app version 17.18 beta reveals a new feature called "Proactive Assistance."

This feature would deliver personalized suggestions based on user activity, drawing data from apps like Gmail and Calendar, on-screen content, and notifications. Privacy safeguards are built in: data processing occurs in a private, encrypted space on the device and is not used for generative AI training or human review.

The existing "Your Day" proactive feed has been renamed to "Daily brief," which may serve as an early implementation of this feature.

Gemini Spark

Branded as "Gemini Spark" (previously "Gemini Agent"), this feature has launched in the United States as of May 2025 for Google AI Ultra subscribers.

Platform Availability:

  • Web: A "Spark" tab appears in the side panel opposite "Chat."
  • Android & iOS: Spark appears between Search chats and Daily brief with a "Beta" label.

Data Access: Integrates with Google Workspace, Connected Apps, Personal Intelligence, signed-in websites, and location. It also includes a remote browser with saved information and a remote computer with code execution data.

Core Components:

  • Task: High-level goal or project
  • Schedule: Specifies when Spark runs (date/time or event-triggered)
  • Skill: Reusable instructions accessed using @ or /

Capabilities:

  • Google Workspace functions: Calendar management, file search and editing in Drive/Docs/Sheets/Slides, Gmail search and composition
  • Declutter inbox: Summarize and archive newsletters, unsubscribe from lists
  • Get meeting briefs: Concise overviews before meetings
  • Get custom news digest: Follow stories over time
  • Deep research: Pull research with cited sources

Task Interface: Includes a "Describe your task" field supporting file uploads and a notebook. Users can take over the remote browser as needed.

Limits: Maximum 15 tasks running simultaneously, subject to the same compute-based limits as other Gemini features.

Remy Agent

Internal documents obtained by Business Insider reveal Google employees are testing "Remy," described as a "24/7 personal agent for work, school, and daily life, powered by Gemini."

Remy is described as "deeply integrated across Google" and capable of monitoring events, handling complex tasks proactively, and learning preferences over time. The upgraded Gemini Agent greeting is: "What can I get done for you today?"

Remy can take actions on the web and with connected apps, such as communicating, sharing documents, and making purchases. The feature remains experimental, with warnings about potential mistakes and data exposure.

Experimental Labs

The Google Gemini web app has introduced an Experimental Labs section to its Tools menu, containing:

  • Agent (available with AI Ultra)
  • Dynamic view or Visual layout (available to all users)
  • Personal Intelligence (available to all paid subscribers)

A "Personalize chat when helpful" toggle has been added, enabling users to manage whether Gemini utilizes their Connected apps for the current conversation.

Gemini Live Upgrades

Beta versions of the Google app (version 17.2) indicate significant upgrades under development for Gemini Live, which currently operates on Gemini 2.5 Flash.

Upcoming Labs features include:

  • Live Thinking Mode: Offers a version that takes more time to process and provide detailed responses, potentially utilizing Thinking or Pro models.
  • Live Experimental Features: Expected to include multimodal memory, improved noise handling, the ability to respond to visual input, and personalized results from integrated Google apps. These are linked to Gemini 3 Flash and Pro.
  • UI Control: Enables the Gemini agent to control the phone to complete specific tasks.
  • Deep Research: Allows users to delegate complex research tasks to Gemini.

Voice Options

Code analysis of Google app version 17.18.22 confirms that existing Gemini voice options are being discontinued.

Strings found in the code state: "Legacy voices are going away" and "Gemini's previous voice options are going away, so you won't see them here anymore."

The affected voices are: Ursa, Nova, Vega, Pegasus, Orion, Eclipse, Capella, Lyra, Dipper, and Orbit.

A removal date has not been specified. Google currently offers 10 English-language voices in the Gemini Android app, including male, female, and a British accent.

Screen Automation & UI Control

Screen Automation

Google is developing a "screen automation" feature for Gemini on Android devices, identified in the Google app 17.4 beta under the codename "bonobo." This aims to allow Gemini to assist with tasks such as placing orders or booking rides within specific applications. Android 16 QPR3 is reportedly laying the groundwork for this integration.

Important: Google has issued advisories stating that users are accountable for actions performed on their behalf and recommends close supervision. Users retain the ability to interrupt Gemini's automation and assume manual control.

Privacy note: Screenshots are subject to review by trained personnel for improving Google services, provided "Keep Activity" is enabled.

AppFunctions

Google has introduced AppFunctions, an Android 16 platform feature and Jetpack library. This allows applications to expose specific functions that agent apps and AI assistants can access and execute directly on the device.

Example use cases:

  • Task management: Creating tasks in productivity apps
  • Media & entertainment: Creating playlists in music apps
  • Cross-app workflows: Using functions from multiple apps
  • Calendar & scheduling: Creating entries without manual interaction

The Samsung Gallery app on the Galaxy S26 and other Samsung devices running OneUI 8.5 and higher utilize AppFunctions. The Gemini app currently employs AppFunctions for its Calendar, Notes, and Tasks integrations.

UI Automation

Google is developing a UI automation framework that enables AI agents to execute tasks on users' installed applications where dedicated app integrations are not yet available. Android 17 is expected to broaden these capabilities. Google is collaborating with a small group of app developers on these experiences.

Avatar Integration

Code analysis of Google app version 17.11.54 reveals development of "Gemini Avatars."

The creation process involves using a phone's camera for a head scan, mirroring the method used for "Likenesses" previously introduced in Android XR for video calls. Users may be able to insert their avatars into Gemini-generated content using prompts such as "@me" or "@%s."

Android Desktop Interface

Google app beta version 17.5 includes strings detailing aspects of the Gemini experience on a desktop version of Android.

Users can access Gemini through a Gemini icon in the top-right corner of the screen or by using a keyboard shortcut involving a Google Key and the Spacebar.

Status & Availability

  • Gemini Spark is currently available in the United States for Google AI Ultra subscribers.
  • Proactive Assistance, Remy, and other features remain unannounced by Google and are identified through code analysis.
  • Google I/O 2025 is scheduled for May 19-20, where the company is expected to showcase new AI products and innovations.

Note: All features identified through APK teardowns and code analysis are subject to change and may not be released to the public.