Google’s Gemini ⁣3 introduces a visually immersive experience, presenting facts ‍in a magazine-style layout with photos ‍and interactive ⁢modules designed to encourage user input and customization. According to Google,⁣ this⁢ design‌ isn’t⁣ merely aesthetic; it’s intended to be highly responsive to user direction.

Introducing Gemini Agent: ‌AI for Multi-Step Tasks

A key component of the Gemini 3 update is the experimental Gemini Agent, a feature enabling the AI to directly handle complex, multi-step tasks within the ​application. This agent can connect to core ⁣Google services – including ‍Google Calendar, Gmail, and Reminders – with user-granted permission to perform actions⁤ like inbox organization and schedule management.

The agent operates by breaking down tasks into manageable steps, providing real-time progress updates, ⁢and crucially, pausing‍ to request user approval before proceeding.Google positions this functionality as ‍a significant ⁣step toward developing a “true generalist agent” capable of handling a wide range‌ of requests.

“Vibe Coding” and Intuitive Task Definition

The approach to task management with Gemini Agent resembles what’s been termed “vibe‌ coding,” where users articulate a desired​ outcome‌ in natural language, and the AI autonomously constructs the necessary interface or code to achieve it. This simplifies complex requests and lowers the barrier to entry for users⁤ without technical expertise.

Deeper integration with Google Search

Gemini ‌3 also strengthens the​ integration ​between Gemini and existing Google products. Select ⁢Google AI Pro and Ultra subscribers ⁣now have access to Gemini 3 Pro,⁤ the model’s reasoning-focused variation, within Google Search. This allows for more in-depth and nuanced AI-generated summaries, leveraging the⁣ model’s reasoning capabilities rather than relying on the standard AI ⁢Mode.

The Gemini Agent will ⁣initially be available on the web to Google AI ultra subscribers in the United States starting November 18, 2023.