Google unveils Jitro
Google introduced Jitro, an AI the company describes as capable of self‑coding and iterative improvement, signaling a push toward more autonomous developer tools. (x.com) The announcement frames Jitro as part of a trend toward AI systems that can refine their own code over time. (x.com)
Google’s new coding push is moving from “write this function” toward “improve this codebase,” with a project called Jitro reportedly positioned as the next step after Jules. (devops.com) Google has not published a standalone Jitro product post on its official blogs, but reports published on April 6 and April 8, 2026 described Jitro as an internal name for a next-generation version of Jules. Those reports said the system is being built around high-level goals, such as better test coverage or lower error rates, instead of one prompt at a time. (testingcatalog.com) (devops.com) To understand that shift, start with what Google already sells. Jules entered public beta on May 20, 2025 as an asynchronous coding agent that connects to repositories, clones code into a Google Cloud virtual machine, and returns a plan, reasoning, and a code diff after it finishes a task. (blog.google) Google expanded Jules on December 10, 2025 with Suggested Tasks, Scheduled Tasks, and a Render integration for “self-healing deployments.” That update let Jules scan code for improvements, automate recurring maintenance, and prepare fixes before a developer manually asked for them. (blog.google) Google then took Jules out of beta on August 6, 2025 and said developers had already produced more than 140,000 publicly shared code improvements during the beta. The company also tied higher usage limits to its Google AI Pro and Google AI Ultra subscriptions. (blog.google) The broader idea is “agentic” software: tools that plan and execute multi-step work instead of only predicting the next line of code. Google has been building that stack across Gemini 3 Pro and Antigravity, a development platform Google introduced in November 2025 and expanded in Google AI Studio on March 18, 2026. (blog.google 1) (blog.google 2) Google’s own Antigravity training materials describe the product as “agent-first” and say autonomous agents can plan, code, validate, and iterate on engineering tasks with minimal human intervention. In a separate codelab updated April 1, 2026, Google said Antigravity acts as a “Mission Control” for managing those agents. (developers.google.com 1) (developers.google.com 2) That puts Jitro in a crowded market. Microsoft’s GitHub Copilot, Cursor, and other coding tools already automate parts of software work, but the reporting on Jitro describes a more persistent system that tracks goals across a codebase instead of waiting for the next instruction. (testingcatalog.com) (devops.com) The missing piece is official detail. As of April 13, 2026, Google’s public materials document Jules, Gemini 3, and Antigravity, but not a formal Jitro launch, pricing page, or product documentation. (blog.google 1) (blog.google 2) (developers.google.com) So the clearest reading for now is narrower than the hype: Google has already shipped coding agents that work in the background, and outside reports suggest Jitro is the company’s next attempt to let those agents choose more of the work themselves. (blog.google) (devops.com)