Zig Creator Slams Anthropic’s Rust Rewrite of Bun, Warns AI Hype Threatens Software Engineering Decisions

Andrew Kelley, the creator of the systems language Zig, has publicly rebuked Anthropic’s recent announcement that the Bun runtime would be ported from Zig to Rust. Anthropic, backed by $132 billion in funding and a projected trillion‑dollar valuation, framed the move as a showcase of AI‑driven code generation. The rewrite was merged without prior notice, prompting a blunt response that highlighted the risks of swapping battle‑tested code for a hype‑driven rewrite. For developers and engineering leaders, the episode illustrates how AI hype can distort language‑choice decisions.

What happened

Anthropic announced that Bun, its JavaScript runtime, would be migrated from Zig to Rust. The company released an explanation two months after the migration landed on the main branch, describing the change as an “unsafe Rust” port that allowed a file‑by‑file migration to minimize risk. The timing and lack of upfront communication let headlines focus on the AI narrative rather than technical trade‑offs.

Zig creator Andrew Kelley responded with a terse statement, calling the move a “spade a spade” and warning that the optics of a high‑profile AI firm rewriting production code could mislead engineers. He emphasized that decisions should be based on clear technical criteria, not on fear of being left behind by AI hype.

Why it matters

The episode highlights three stakes for engineering teams: the loss of battle‑tested Zig code that has accumulated real‑world reliability, the hidden cost of a large‑scale language migration, and the broader narrative that AI can replace human judgment in architectural choices. When funding giants like Anthropic promote AI‑generated rewrites, product managers and CTOs may feel pressure to adopt the latest hype, potentially sacrificing stability and long‑term maintainability.

How to think about it

1. Measure, don’t assume – Benchmark the current Zig implementation against a prototype in Rust before committing.
2. Incremental migration – Move isolated modules behind feature flags rather than a wholesale rewrite.
3. Team expertise – Align language choice with the existing skill set; invest in training only if the benefits outweigh the cost.
4. Separate AI hype from engineering merit – Treat AI‑generated suggestions as data points, not directives.
5. Post‑migration validation – Require real‑world usage and feedback before deprecating the original code.

FAQ