News

Google Counters GitHub & Microsoft with Jules Agent & Enhanced Gemini AI

• By David Ramel
• 05/20/2025

As our new age of agentic AI takes shape, Google is stepping up its game with the introduction of the Jules agent and enhanced Gemini AI, which could be seen as a direct response to the growing competition from Microsoft and GitHub, who have been making significant strides in the space.

Google made several AI-related announcements today at the start of its annual I/O developer conference, including the introduction of the Jules agent. This new AI agent is designed to assist users in various tasks, leveraging the power of Gemini AI to provide enhanced capabilities. Also, the company's flagship AI model, Gemini, has received significant updates, making it more powerful and versatile.

Google's moves follow this week's announcement from OpenAI that Codex AI is returning in a new form from that model that powered the original GitHub Copilot, now described as "a cloud-based software engineering agent," along with a slew of agentic AI initiatives announced just this week by Microsoft at its Build 2025 developer conference, where it heralded the "age of AI agents" [Editor's note: This sentence has been corrected to specify that the announcement came from OpenAI, not GitHub, as was originally mistakenly reported].

While the Gemini ecosystem might be seen as the company's main agentic AI offering to counter rival alternatives like Codex and all the Copilots, the new Jules agent has a more specific purpose.

Jules documentation indicates it's an experimental GitHub-integrated AI coding agent that asynchronously fixes bugs, adds documentation, and builds new features, previously available only in a limited preview.

"Now available to everyone, Jules is an asynchronous coding agent that gets out of your way, so you can focus on the coding you want to do, while Jules picks up the random tasks that you'd rather not," the company said today of the offering just emerging from a limited preview. "It can tackle your backlog of bugs, handle multiple tasks at once, and even take the first cut at building out a new feature. Jules works directly with GitHub, clones your repository to a Cloud VM and when you're ready, creates a PR that you can merge back into your project."

Highlights include:

• Uses Google's Gemini 2.5 Pro multimodal AI model optimized for coding
• Breaks down complex assignments into multi-step plans and adapts to user instructions
• Runs unit tests to verify code changes before submitting pull requests
• Provides audio summaries of pull requests to explain modifications, inspired by Google's NotebookLM feature
• Offers real-time progress updates and notifications through browser alerts or the Jules interface
• Enables task prioritization and status adjustments via plugins or notification systems
• Supports customization for specific project environments using setup scripts
• Supports Python and JavaScript projects initially, with plans to expand language support
• Includes a free tier with five tasks per day and two concurrent tasks, making it accessible for individual developers

Speaking of Gemini Pro 2.5, it received significant updates a few weeks ago, and more new features and functionality were announced today for that model and the company's AI portfolio.

Google unveiled a suite of new AI tools, models, and platform enhancements designed to empower developers and accelerate the creation of advanced AI applications. Major AI highlights announced today include the debut of Gemini 2.5 Flash for faster, more efficient coding and reasoning, and the introduction of several new models tailored for specific use cases -- such as Gemini Diffusion for rapid text generation, Gemma 3n for multimodal tasks on edge devices, and Lyria RealTime for interactive music creation. Google also expanded its Gemma family with MedGemma for medical applications and SignGemma for sign language translation. Developers benefit from a revamped Google AI Studio, new agentic features in Colab, the general availability of Gemini Code Assist, and the launch of Firebase Studio for streamlined app development. Additional tools like Stitch for UI design, enhanced Gemini API capabilities (including asynchronous function calls, computer use, and URL context), and support for the Model Context Protocol round out a comprehensive set of updates aimed at making AI development more accessible, powerful, and versatile.

Specific, fleshed-out highlights include:

• Gemini 2.5 Flash: New version optimized for speed and efficiency, with improved coding and complex reasoning performance. Available in Google AI Studio and Vertex AI in Preview; general availability coming soon.
• Thought Summaries & Thinking Budgets: Summaries now available across Gemini 2.5 models; thinking budgets coming to 2.5 Pro to help manage costs and model behavior.
• Gemini Diffusion: State-of-the-art text model delivering outputs 4-5 times faster than comparable models, open to trusted testers.
• Gemma 3n: Fast, efficient open multimodal model for text, audio, image, and video, designed to run on phones, laptops, and tablets.
• Lyria RealTime: Experimental interactive music generation model for real-time music creation and performance via Gemini API and Google AI Studio.
• MedGemma: Multimodal medical text and image model for health applications, now available as part of Health AI Developer Foundations.
• SignGemma: Upcoming open model for translating sign language into spoken language text, enabling new accessibility apps.
• Google AI Studio Updates: Cleaner UI, integrated documentation, usage dashboards, new apps, and a Generate Media tab for generative model experimentation.
• Agentic Colab: Soon to offer a fully agentic experience -- users can describe goals and watch Colab take actions, fix errors, and transform code automatically.
• Gemini Code Assist: Now generally available for individuals and GitHub users, with Gemini 2.5 powering the assistant and a 2 million token context window coming soon. Of special interest to readers of Visual Studio Magazine, Gemini Code Assist is available as a Visual Studio Code extension boasting more than 700,000 installs.
• Firebase Studio: New cloud-based AI workspace for building full-stack AI apps, including Figma integration and automatic backend provisioning.
• Stitch: AI-powered tool for generating UI designs and frontend code from natural language or image prompts, with easy export to CSS/HTML or Figma.
• Gemini API Enhancements:
  • Native audio output and live API features, including proactive video/audio and affective dialog support.
  • Asynchronous function calling for background execution of long-running tools or functions.
  • Computer Use API for building apps that can browse the web or use software tools on behalf of users (rolling out to more developers soon).
  • URL Context tool for retrieving in-depth content from URLs, supporting research agent development.
  • Model Context Protocol (MCP) support for easier integration with open source tools via the Gemini API and SDK.

Google I/O 2025 concludes tomorrow. CEO Sundar Pichai's keynote address is covered in this post.