Google has officially launched Jules, a new AI-powered coding assistant designed to autonomously perform coding tasks for developers. The Joule, which was first introduced last December and has been available in beta since then, is now available in the public eye. The assistant is powered by Gemini 2.5 Pro, Google's latest multimodal leading language model, brings advanced inference and planning capabilities to workflow coding.
According to Google, Joule is a new AI-powered coding assistant built to work asynchronously. Unlike typical coding tools that run in real time within a limited sandbox, Joule solidly clones the entire developer's codebase into a Google Cloud Virtual Machine (VM). From there, you can perform tasks independently, such as creating tests, fixing bugs, building features, updating dependencies, and even generating audio change logs.
According to Google, this asynchronous operation allows developers to assign tasks to AI agents and focus on other tasks. Once the task is complete, the AI agent provides developers with a comprehensive summary that includes detailed differences in their plans, inferences, and code changes. Meanwhile, the company also emphasizes that Julie offers stealability. This means that developers can review and coordinate AI proposal plans at any stage, giving them complete control over the process.
Meanwhile, Joule is also said to be designed to integrate directly with existing repositories, including seamless Github support. This allows developers to stay within their familiar workflows without the need for additional setups or tool switching. Furthermore, according to Google, Jules is optimized for efficiency with the ability to run multiple coding tasks in parallel within a secure cloud VM. This reduces turnaround times for complex or large codebases.
Google reports that in the beta stage, thousands of developers have used joules to complete tens of thousands of tasks, bringing more than 140,000 public code improvements. Based on feedback, the company has significantly upgraded to AI agents, including sophisticated interfaces, bug fixes and additional features. These new features include faster execution, multimodal input support, and deeper GitHub publishing integration, reusing previous setups.
The company also assures users that the assistant is private by default. This means that it doesn't train with the user's private code. Additionally, all code and data remains isolated within the secure execution environment provided by the VM infrastructure. Google emphasizes that privacy and security are core design considerations throughout development.
Joule Planning and Availability
Google has announced the general availability of Joule AI with three structured access layers:
Introductory access to Joules: This tier targets new users, allows you to explore Joule's core features.
Joules in Google AI Pro: Suitable for daily coding tasks, this layer offers five times the usage limits of the introductory level.
Google's Joule AI Ultra: Designed for rugged multi-agent workflows, this layer offers 20 times the limit of the base layer.
These new plans are soon rolling out to Google AI Pro and Ultra subscribers. In particular, eligible university students can receive a free one-year subscription to AI Pro Tier. This can provide valuable resources for both learning and software development.
– end
