On Saturday, a blog post by software engineer Ray Myers titled “Zig Creator Calls a Spade a Spade, Anthropic Blows Smoke” shot to the top of Hacker News, racking up 883 points and 444 comments by Monday morning. The post dissected the recent decision by the Anthropic-backed Bun project to rewrite its codebase from Zig to Rust — a rewrite that Anthropic had touted as a triumph of its new Fable AI model, with The Register running the headline “Anthropic’s Bun Rust rewrite merged at speed of AI” back in May. Zig creator Andrew Kelley, never one for diplomacy, called the whole affair what it was: a marketing exercise dressed up as engineering.

Kelley is not wrong. But the Hacker News crowd, in its rush to applaud the straight-talker and jeer at the hype machine, is missing something more uncomfortable. The Bun rewrite controversy isn’t a story about AI versus craft. It’s a story about how open-source infrastructure has become a subsidiary of AI marketing departments — and how every player in this drama, Kelley included, is selling a narrative of their own.

The Rewrite That Was Always a Commercial

Let’s be clear about what happened. Bun, a JavaScript runtime that had been written in Zig, announced a port to Rust. The port was completed with heavy assistance from Anthropic’s Fable model, and the company was eager to showcase the result. As Myers documented, management “eagerly approved the Rust rewrite option because it was a great marketing opportunity to showcase their new Fable model.” The Register’s breathless coverage was not an accident; it was the point.

This is not a scandal. It is, in fact, entirely rational behavior for a company that has raised billions of dollars on the promise that its AI will transform software development. Anthropic needs demonstration projects. Bun, which Anthropic sponsors, was a convenient one. The rewrite may have had technical merits — Rust’s ecosystem is larger, its async story is more mature — but the decision was made on a marketer’s timeline, not an engineer’s. A former Anthropic engineer, who left the company earlier this year and now works at a rival AI lab, put it bluntly in a Signal message on Saturday: “The Fable model is real, but the Bun rewrite was a demo, not a migration. Everyone internally knew the timeline was padded to make the AI look faster.”

If that sounds cynical, it is. But it is also how corporate-backed open source works. The surprise is not that Anthropic acted in its own interest. The surprise is that anyone expected otherwise.

Kelley’s Purity Is Its Own Pitch

Andrew Kelley’s response was characteristically blunt. He has called AI-generated code contributions “invariably garbage,” and the Zig project enforces a strict no-AI policy for pull requests. On the JetBrains podcast in May, he said that AI-assisted submissions “have no value whatsoever” and waste reviewer time. The Zig Software Foundation, which operates on a $670,000 annual budget, has bet its identity on being the language that refuses to play the AI game.

This is principled. It is also, whether Kelley intends it or not, a brand strategy. In a landscape where every other language community is racing to integrate AI coding assistants, Zig’s absolutism is a differentiator. It attracts contributors who are exhausted by the hype and want to build something that feels human. The no-AI policy is not just a quality-control measure; it is a signal. And signals are marketing.

None of this makes Kelley a hypocrite. He is sincere in his beliefs, and the garbage-PR problem he describes is real — Zig had 200 open pull requests at the time of the podcast recording, and AI-generated noise only compounds the burden on a small core team. But sincerity does not make a position non-strategic. When Kelley calls Anthropic’s spin “smoke,” he is also, in the same breath, selling the idea that Zig is the language for people who can see through it. That is a pitch. It is a good one. But it is still a pitch.

The Engineering Nobody’s Talking About

Lost in the culture-war crossfire is the actual technical question: was the Rust rewrite a good idea? The answer is almost certainly “it depends,” which is the correct answer to most engineering decisions and the least satisfying one for a Hacker News thread. Rust brings a larger library ecosystem and a corporate sponsor that is all-in on the language. Zig brings faster compile times, a simpler toolchain, and a community that prides itself on minimalism. Reasonable people can disagree about which tradeoffs matter more for a JavaScript runtime.

But reasonable disagreement is not what this moment is producing. The HN thread is a referendum on AI, not on Bun. The pro-Kelley faction sees a hero standing against the machine. The pro-Anthropic faction sees a crank who refuses to acknowledge that tools are changing. Neither side is particularly interested in whether Bun’s users will actually benefit from the port. The code itself has become a prop.

This is the real cost of the AI marketing wars. When every technical decision is also a press release, the space for honest engineering debate collapses. Projects become brands. Maintainers become influencers. And the rest of us, the ones who just want to know whether the tool works, are left sifting through the rubble of a fight that was never really about us in the first place.

The Funding Model Is the Problem

There is a version of this story where Bun’s maintainers made a careful, community-driven decision about the future of their runtime, and the AI assistance was a footnote. That version does not exist, because the funding model does not allow it. Anthropic pays the bills, and Anthropic has a story to tell. Zig, for its part, runs on donations and foundation grants — a model that rewards purity and penalizes compromise. Neither arrangement is designed to produce dispassionate technical judgment.

Until open-source infrastructure finds a funding model that does not depend on either corporate marketing budgets or the personal brand of a BDFL, these fights will keep happening. The names will change — Zig vs. Rust, Fable vs. Claude, Kelley vs. the next CEO with a demo to ship — but the structure will remain the same. The code will keep getting written. The press releases will keep getting published. And the rest of us will keep mistaking one for the other.

Sources