Олег Онопрієнко AI Eng 7 August 2025, 08:30

Google's AI agent for vibecoding Jules has exited beta testing

During testing, Google Labs added new features, including reusing previous settings to complete tasks faster, integration with GitHub issues, and support for multimodal input.

8

Google's AI agent for vibecoding Jules has exited beta testing

During testing, Google Labs added new features, including reusing previous settings to complete tasks faster, integration with GitHub issues, and support for multimodal input.

After two months of closed testing , Google has released the public version of its asynchronous AI programming assistant, Jules. The AI assistant was first introduced in December 2024 as a “trusted automated programming assistant” that can handle a wide range of time-consuming tasks on behalf of users.

Jules is based on the large Gemini 2.5 Pro language model, which Google introduced in March as “the most advanced model for complex tasks.”

By running asynchronously in a virtual machine, Jules differs from leading AI coding tools such as Cursor, Windsurf, and Lovable, which run synchronously and require users to monitor the results after each prompt.

“Jules works like an extra pair of hands… You can give it a task and then close your computer and come back a few hours later. Jules will complete those tasks for you, unlike if you were doing it with a local agent or using a synchronous agent, you would be tied to that session,” Google Labs product director Katie Korewets explained to TechCrunch.

During the beta, thousands of developers tackled tens of thousands of tasks, resulting in over 140,000 code improvements. With developer feedback, Google improved the user interface, fixed hundreds of bugs, and launched new features, including reusing previous settings to make new tasks run faster, integration with GitHub, and support for multimodal tasks.

Subscribers to Google AI Pro ($19.99) will get five times the limits for Jules, while Google AI Ultra ($124.99) will get a twenty-fold increase over the basic version.

Google has also updated Jules’ privacy policy to more clearly explain how it trains the AI. If the repository is public, its data can be used for training, but if it’s private, no data is sent.

Previously, Jeff Dean, Google's chief scientist, said that artificial intelligence could soon reach the level of skills of a junior software engineer . He believes that such an AI should know programming better than at a basic level.