Skip to content
Theme
OnDevice Tools

SoftwareOpen Source & Linux

GitHub turns AI bug reports into fixes for open source projects

A GitHub and Microsoft pilot called Codename MDASH converted thousands of raw AI findings into 99 checked private reports with suggested fixes. Maintainers have already repaired more than half of them.

Illustration of an open source code repository with an AI report and a fix

Key takeaways

  • GitHub and Microsoft ran the Codename MDASH pilot on six widely used open source projects.
  • Thousands of raw AI findings became 99 checked reports, each filed privately with a suggested fix.
  • GitHub says maintainers have already fixed more than half of those 99 reports.
  • GitHub also now highlights ACCESSIBILITY.md files for documenting accessibility goals and barriers.
In this article
  1. A pilot called Codename MDASH
  2. More than half already repaired
  3. Accessibility files get highlighted
  4. Why it matters for open source

A pilot called Codename MDASH

GitHub and Microsoft ran a pilot named Codename MDASH across six widely used open source projects, according to the GitHub Blog. The work started with thousands of raw findings produced by AI tools. Those findings were then sorted and checked, leaving 99 reports that were filed privately and each came with a suggested fix.

More than half already repaired

GitHub says maintainers have already fixed more than half of the 99 reports. The company published its account on 17 November 2026 under the argument that maintainers need patches rather than a larger pile of AI bug reports. In GitHub's framing, an automated finding only helps a project when someone also supplies the repair.

Accessibility files get highlighted

The same GitHub Blog material notes a separate change for project documentation. GitHub now highlights ACCESSIBILITY.md files, which maintainers can use to record accessibility goals, known barriers, supported environments, contributor expectations and how problems should be reported.

Why it matters for open source

Illustration: code, GitHub logo, AI, bug and the labels AI, Codename Mdash, GitHub

Both items reduce work for the people who keep widely used projects running. Reviewing AI output takes time, and a checked report with a patch attached is cheaper to act on than an unverified alert. The published material does not say whether the MDASH pilot will be extended beyond the six projects it covered.

Sources

  1. Maintainers don’t need more AI bug reports. They need fixes.GitHub Blog

Tags

Spotted a mistake? See our corrections policy. Read how News by OnDevice works in our editorial policy.