Skip to content- Watching for: Bun 1.4.0 stable past the current canary; regression reports from the Rust port; independent confirmation of the 2-5% throughput / ~20% binary-size / leak-fix claims; which model actually drove it (blog says Claude Code, HN attributes Fable 5).
Notes
- 2026-07-09: Bun published 2026-07-08 an account of rewriting ~535k LOC (transpiler, package manager, test runner, Node APIs) from Zig to Rust, motivated by memory-safety bugs from mixing GC'd JS values with manual memory. Work ran May 3-14 2026 with many parallel Claude Code instances (~6,500 commits, up to 64 at peak) using adversarial review loops. Claims 2-5% higher throughput, ~20% smaller binaries (Linux/Windows), fixed leaks. Merged to main, ships in Bun 1.4.0 canary; Bun states Claude Code v2.1.181+ already use the Rust port; 1.3.14 was the last Zig release. Covered 2026-07-09 Top stories. Part of the Fable 5 capability-demo wave. Vendor claims, unverified. 2026-07-09: Zig creator Andrew Kelley published a rebuttal (andrewkelley.me/post/my-thoughts-bun-rust-rewrite.html, HN 48843352) arguing the gains were not from the language switch (Zig supported LTO throughout but Bun kept it disabled over LLVM bugs that also affect Rust; Zig ships comptime/inline audit tooling Bun did not use), that eliminating memory-safety bugs is mainly engineering effort not language choice, and disputing the post's fuzzing claim as a fabrication; also flags low-quality AI contributions to Zig. One maintainer's assertions, unverified. Covered 2026-07-09 Reddit and social pulse.
Stories