TLDRocket
Sign in

GitHub and Anthropic used their own agents for major Rust rewrites — with very different playbooks

The New Stack Paul Sawers

GitHub rewrote Copilot’s runtime in Rust using its own AI agents. Anthropic did something similar with Bun, but took a very different route.

Based on reporting by The New Stack, Paul Sawers — read the original for the full story.

Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error

Rust keeps showing up wherever companies want safer, steadier infrastructure, and GitHub just put its own money where its mouth is. The company says it has rewritten the GitHub Copilot agent runtime, which had lived in TypeScript on Node.js and V8, into more than 800,000 lines of production Rust. The twist is the one that matters: GitHub says the migration was done with its own Copilot coding agents, including the Copilot app and Copilot CLI.

Microsoft engineer Stephen Toub says the work ran through 128 pull requests and was shipped in stages, not in one giant flag day. That slower approach let the team catch regressions as they appeared instead of discovering them after the fact. Toub’s pitch is blunt: a rewrite that would once have tied up a whole team for a year or two was finished mostly by a single developer in a few months, while the rest of the team kept extending the runtime.

Anthropic’s Bun story reaches the same headline from a different angle. Bun, which Anthropic acquired in December, announced in July that it was moving more than half a million lines of Zig into Rust, with founder Jarred Sumner leaning on a pre-release Claude model. Bun had grown from a 2021 project built in a cramped Oakland apartment into a runtime whose CLI gets more than 22 million monthly downloads, but it was also dealing with memory-management bugs, including leaks and crashes.

Sumner’s answer was not to keep patching each bug one by one forever. He said that would have been a year-long job for a small team and would have frozen bug fixes, security work and new features. Instead, multiple Claude Code agents translated and checked different parts of the code at once while he supervised, and after 11 days the Rust version was already passing Bun’s existing tests on all six of its platforms.

The comparison is less about which agent is smarter and more about which migration strategy fits the job. GitHub went incremental and kept shipping. Bun went wide and fast, then cleaned up after. OpenAI has now added its own quiet example too, with Habitat rewritten from Python to Rust and already handling 95% of production requests, which suggests this is becoming a pattern rather than a stunt.

My take — AI-written commentary, not fact-checked reporting

The real headline isn’t that agents can help with Rust rewrites. It’s that the people selling those agents are now using them as proof of concept, which is very convenient and only slightly self-serving. Still, if the code compiles, passes tests, and ships without turning the team into a bonfire, the market will happily call that progress.

Read more about this at: The New Stack

Related stories

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.