GitHub rewrites Copilot runtime in Rust using AI agents
A project expected to take a year or two was completed in months by a single developer, with AI writing most of the 800,000 lines of code.
What to know
- GitHub migrated the Copilot agent runtime from 800,000 lines of TypeScript to production Rust, with AI agents writing most of the code across 128 pull requests.
- The project was completed in a few months by a single developer—work that would have required a full team for 1–2 years before AI agents became available.
- The rewritten Rust runtime achieved orders-of-magnitude performance improvements and shipped incrementally with regressions caught and fixed along the way.
- The runtime backs Copilot CLI, the Copilot app, Copilot SDK, and a growing set of Microsoft and GitHub ecosystem AI solutions.
GitHub Developer platform company
How it unfolded 1 development · click the chart to see its coverage articlesposts
-
1
Copilot runtime performance improves by orders of magnitude
The rewritten Rust runtime achieved substantial performance gains compared to the original TypeScript implementation on Node.js. The runtime now backs the Copilot CLI, Copilot app, Copilot SDK, and a growing set of Microsoft and GitHub ecosystem solutions.
“the performance of the runtime improved by orders of magnitude…”
— GitHub Blog -
first by GitHub Blog, 10d ago
-
-
background
AI agents write most of Copilot runtime in 128 pull requests — GitHub reports that AI agents wrote most of the 800,000 lines of Rust code across 128 pull requests that landed in main. The code was shipped incrementally rather than in a single cutover, with regressions discovered and fixed quickly. A project expected to take a full team 1–2 years was completed primarily by a single developer in a few months.
-
background
GitHub announces Copilot runtime migration to Rust — GitHub published a blog post detailing the completion of a rewrite of the Copilot agent runtime from TypeScript to Rust, spanning over 800,000 lines of production code. The post emphasizes that the project scale would not have been economically feasible before the availability of AI agents.