Microsoft agentically ports Copilot runtime to Rust for $120K
Microsoft rewrote its GitHub Copilot runtime from TypeScript to Rust using AI agents, converting 430,000 lines for about $120,000 in tokens.
Microsoft ported the runtime underpinning GitHub Copilot from TypeScript to Rust, with AI agents running GPT-5.6 Sol and Claude Opus 4.8 performing most of the conversion across 135 releases over 14.5 weeks. The project converted 430,000 lines of TypeScript into 800,000 lines of Rust for roughly $120,000 in token costs plus about three weeks of developer time, delivering a 15.9x throughput speedup and cutting memory from 1,383 MB to 126 MB for a 10-client agent batch. Managers had to fix a few dozen regressions, and the port used an iterative investigate-hypothesize-change loop, including a 25-hour session on a 30,000-line session.ts file that spawned 15 coordinated child sessions.
- Agents converted 430,000 TypeScript lines into 800,000 Rust lines across 135 releases.
- Rust delivered a 15.9x throughput speedup and roughly 91% memory reduction in benchmarks.
- GPT-5.6 Sol and Claude Opus 4.8 handled different parts of the port based on strengths.
- Dozens of compiler-approved regressions highlight limits of agentic code migration.
- Runtime backs GitHub Copilot CLI, VS Code, Visual Studio, Excel, Outlook, and PowerPoint.
- Products
- GitHub CopilotVS Code
- Organizations
- Microsoft
- AI models
- GPT-5.6 SolClaude Opus 4.8
From the original article
4 min read at theregister.com
The software engine underpinning GitHub Copilot and a growing number of Microsoft products is now written entirely in Rust, with AI agents doing most of the porting work. The migration cost about $120,000 in AI token usage plus about three weeks of a developer's time. However, managers also had to grapple with a few dozen regressions in the resulting code, pointing to AI’s ongoing challenges in understanding Rust.
By pjmlp · theregister.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.