April 2, 2026
Speed hype, font gripe
We sped up bun by 100x
100x faster Bun claim sparks cost questions, nitpicks, and a font feud
TLDR: A team says they made the Bun web tool 100x faster by rewriting a code tracker in Zig and using AI-powered swarms, with a flashy browser demo. Commenters question the true price tag, challenge a mislabel about libgit2, and note incomplete tests, splitting the crowd between wow and wait-a-minute.
The team behind Bun says they turbocharged it by 100x after rewriting the code tool Git in Zig and unleashing an AI “swarm” of coding agents. They even brag 4x faster than Git on a Mac and a browser-friendly demo thanks to WebAssembly (a way to run fast code on the web). There’s a demo to clone repos and a whole spin-off thread to debate the Zig-based Git client on HN.
Cue the crowd: the loudest chorus is “show us the receipts.” One commenter ran “napkin math” on what Claude (the AI helper) might’ve cost and tossed out $67,000. Another nitpick turned into a fact-check brawl: the post calls libgit2 “git’s C library,” and critics fired back that it’s not officially part of Git at all. Then the stickler squad pointed to the fine print: ZigGit didn’t finish Git’s enormous test suite, so the “drop-in replacement” claim feels… aspirational. Meanwhile, the peanut gallery launched a side war over the site’s eye-searing font—because of course.
The vibe is peak internet: huge speed flex, murky cost, questionable wording, and a font fight nobody asked for. Fans cheer the browser demo and raw performance; skeptics want full tests, clearer benchmarks, and a real price tag. TL;DR: a rocket boost, with a lot of smoke.
Key Points
- •The team rewrote Git in Zig, reporting 100x faster Bun operations.
- •Their Zig-based Git client is claimed to be 4x faster than standard Git on an ARM MacBook.
- •Compiling the Zig Git implementation to WebAssembly yielded a 5x smaller build with 8.5x more exports.
- •A browser demo is available to clone a repository using the WASM build of the Zig Git client.
- •They detail a ‘code cannon’ setup using the vers CLI, environment variables (GitHub, vers, Anthropic), plan-driven agents, and VM-based execution with periodic human check-ins.