Zig Creator Calls Bun's Rust Rewrite Unreviewed Slop

Bun rewrote 535,496 lines of Zig into Rust in 11 days using about 64 concurrent Claude Fable 5 agents, per the company's own blog. If a passing test suite really can replace human review at this scale, every team doing large AI rewrites has to rethink its risk model. Bun's bet cost about $165,000 in API spend and produced a half-million-line codebase no engineer there has read end to end.

The agents hit 1,300 lines a minute at peak, spread across 6,778 commits. Every test passed on all six platforms, Bun says.

Andrew Kelley, who created Zig, called the result "unreviewed slop," per The Register. Kelley's pushback: "The argument for shipping all the million lines of unreviewed code is that the test suite is good enough to catch everything." His point: if the suite couldn't catch every bug in the original Zig, trusting it alone on a million unreviewed lines is a stretch.

Yes, but the counter isn't weak: a port is mechanical, semantics are pinned on both ends, and Rust's type system catches bugs Zig's doesn't.

Here's what a passing test suite actually proves once the same agents wrote the code and clear the bar. It stops being an independent check and starts being a mirror of the code it's grading.

A Fireship video on the rewrite topped 397,000 views within a day of its July 15 upload. It called the port "the most controversial rewrite in history."