The latest upgrade to the Jules agent introduces several behind-the-scenes improvements aimed at developers working with AI-driven automation in code repositories. Jules now leverages a new ability to read from an AGENTS.md
file if it’s present, giving the agent more relevant context about project-specific workflows and instructions. This change enables Jules to align its actions more closely with each repository’s unique setup and requirements.
New Jules agent just dropped.
— Jules (@julesagent) June 20, 2025
- Reads https://t.co/RuQgBubWQL
- Faster runs
- Less punting when it can actually do the thing
- Reliable env setup (if you’ve scripted it)
- Improved test writing & running
Try it out and tell us what you break. Follow along with the full… pic.twitter.com/JWuEOlv14Z
Performance enhancements are a core part of this update, with users expected to notice faster task completion. While the team hasn’t released specific benchmarks, the improvement is intended to make the agent’s operation feel more responsive during typical development cycles. Another notable change is a reduction in “punting”, the tendency for the agent to give up on tasks or require manual intervention. The updated logic helps Jules push through more tasks independently, providing a smoother user experience.

Reliability during setup processes has also been improved. If a project includes an environment setup script, Jules will now run it more consistently, addressing previous gaps where setup could fail or be skipped. The agent has also adopted better practices around testing: it’s now more likely to write and execute tests automatically, encouraging better code quality within automated workflows.
Jules is an AI agent focused on automating software development processes, particularly in environments where repositories are complex or have specific configuration needs. The recent enhancements further the project’s goal of making intelligent automation accessible and dependable for a wider developer audience.