By ByteBot
Anthropic is rewriting Bun—the JavaScript runtime with 89,000 GitHub stars and 7 million monthly downloads—from Zig to Rust using Claude AI to auto-generate the port. A GitHub branch named claude/phase-a-port shows 1,799 files changed across 43 commits, and the comparison is so large GitHub can’t even render it. This isn’t a refactor. It’s a complete language migration driven by AI, and it’s happening right now.
The industry is about to find out whether AI can reliably rewrite a production runtime used by millions of developers. The implications go far beyond Bun.
The “Vibe Coding” Controversy
Here’s where it gets interesting. The rewrite isn’t just about Rust being technically superior to Zig. It’s about which language Claude can write better. Rust has vastly more training data in LLM datasets than Zig, making Claude significantly better at generating Rust code. The Hacker News community has a term for this: “vibe coding”—making architectural decisions based on AI tooling capabilities rather than pure technical merit.
The Hacker News thread has 236 comments and counting, with developers split down the middle. Some see this as pragmatic modernization. Others see it as a “massive undertaking for vibe coding,” questioning whether a company should rewrite an entire production system based on which language their AI handles better.
But here’s the thing: AI capabilities are a legitimate technical consideration in 2026. When your development velocity depends on AI-assisted coding, the language you choose matters. If Claude can port 1,799 files of systems code from Zig to Rust reliably, that’s not “vibes”—that’s a measurable advantage. The question is whether it can actually pull that off without introducing subtle bugs.
Zig’s Anti-AI Policy Forced Their Hand
There’s more to this story. Bun achieved a 4x performance improvement through parallel semantic analysis and LLVM backend optimizations. The problem? They couldn’t upstream the changes to Zig because the Zig project has one of the strictest anti-AI policies in open source—no LLM-generated code in pull requests or even bug tracker comments.
Zig’s rationale is philosophical. They value “contributors over contributions,” viewing each contributor as an investment in long-term relationships. They call it “contributor poker”—the relationship with the developer matters more than the code itself. AI-generated PRs don’t build those relationships, so they’re banned outright.
It’s a principled stance, but it comes with real costs. Bun faced a choice: fork the Zig compiler and maintain it indefinitely (ongoing tech debt), or switch languages entirely. Anthropic chose the latter. Zig’s purity just cost them their highest-profile project, and the timing—right after the AI policy clash—tells the story.
The Stakes Are Massive
This isn’t a hobbyist experiment. Bun has 7+ million monthly downloads and is used by companies like Midjourney and Cursor. An AI-generated rewrite of this scale risks introducing bugs that could ripple through millions of production deployments. The HN community is asking the right questions: Can AI reliably port 1,799 files of systems-level code? What about edge cases? What about testing coverage?
Anthropic acquired Bun in December 2024 as part of their Claude Code infrastructure strategy. Claude Code hit $1 billion in revenue just six months after launch, with clients including Netflix, Spotify, and Salesforce. If this rewrite succeeds, it’s a massive validation of AI-driven development at scale—and a marketing win for Anthropic. If it fails, it becomes a cautionary tale that reinforces skepticism about AI code quality in production systems.
The entire developer community is watching.
What This Means for Language Adoption
This signals a fundamental shift in how languages get adopted. In 2026, “LLM training data availability” is becoming a technical consideration alongside performance, safety, and ecosystem maturity. Rust is winning not just because of its borrow checker and safety guarantees, but because there’s enough Rust code in training datasets for LLMs to write it competently.
Pre-1.0 languages and niche languages now face a new adoption barrier: insufficient training data for AI tools. Zig losing Bun—its most visible project—damages the ecosystem’s credibility. Meanwhile, Rust continues its trajectory of replacing C and C++ in critical infrastructure. It’s already in the Linux kernel, Windows components, and AWS services. Now it’s absorbing JavaScript runtimes too.
The vibe coding trend explains why. TypeScript, Python, and Rust dominate “AI-friendly language” discussions because they’re well-represented in training data. Developers optimizing for AI-assisted workflows will gravitate toward these languages, creating a self-reinforcing cycle.
The Verdict Is Still Out
Whether this rewrite succeeds or fails will set a precedent. If Claude can port a production runtime without introducing critical bugs, expect more companies to attempt AI-driven rewrites. If it stumbles, it validates the skeptics who argue that systems-level code is too complex for current AI capabilities.
Either way, the decision to prioritize AI tooling over language purity reveals where the industry is heading. Zig’s anti-AI stance is principled, but Anthropic’s bet on Rust and Claude is pragmatic. One approach values contributor relationships; the other values development velocity.
The Bun rewrite is the experiment that will tell us which philosophy wins.
