Claude Fable 5 rewrote Bun JavaScript runtime in Rust with over 1 million lines of code in 11 days

Claude Fable 5 helped rewrite the Bun JavaScript runtime from Zig to Rust in eleven days.

Jarred Sumner, Bun creator, ran 64 model instances in parallel. They produced more than one million lines of code. The API bill reached roughly 165000 dollars. The move replaced a language that produced repeated memory errors with one that catches those errors at compile time.

Bun v1.4.0 arrived as a Canary release. It closed 128 bugs and posted two to five percent gains in speed on core workloads.

Why the rewrite happened now

Bun started in Zig for its simplicity and speed. Memory safety bugs still appeared often enough to threaten stability. Rust offered the compiler checks Sumner needed without changing the performance target.

The Anthropic acquisition of the Bun team in December 2025 added resources and pressure to ship cleaner infrastructure. Model scale made the parallel rewrite feasible.

How 64 instances worked at once

Sumner split the codebase into narrow tasks. Each Claude Fable 5 instance owned one module or subsystem. A central coordinator merged changes and ran tests every few hours.

Compile-time guarantees caught dangling pointers and data races that had required manual Zig debugging. The process ran twenty-four hours a day for the full eleven days.

What the new release delivered

Version 1.4.0 Canary removed the largest class of runtime crashes reported in prior months. Users testing the build saw modest latency drops on HTTP and file-system heavy scripts. No new breaking changes appeared in the public API.

The team kept the same binary size and kept startup time within one percent of the previous Zig build.

Limits that remain after the switch

Memory safety improved at the cost of longer compile times during development. Some low-level optimizations still sit behind unsafe blocks that the team must audit manually. The 165000 dollar experiment cost will not repeat for every future change.

Competitors using Go or C++ for their runtimes have not announced similar model-driven rewrites. Whether the cost curve falls fast enough for others to copy remains open.

What to watch next

The first stable release after Canary will show whether the speed gains hold under broader workloads. Anthropic could publish internal metrics on how often developers now choose Rust after this case. Any new memory-related bug reports in the next quarter will test if the language switch delivered lasting reliability.