Zig Creator Andrew Kelley Calls Out Anthropic Over Bun Rust Rewrite Narrative

The Bun project's decision to rewrite from Zig to Rust has sparked one of the most heated debates in the programming community this year. Zig creator Andrew Kelley published a blunt response to Bun's explanation. The fallout is forcing hard conversations about AI-generated code, engineering culture, and how tech narratives are shaped by massive funding.

The Core Disagreement

Two weeks ago, Bun (the JavaScript runtime built by Oven, an Anthropic-backed startup) published a detailed blog post explaining why they were porting their codebase from Zig to Rust. They cited memory management bugs and the complexity of building a production infrastructure project in a systems language still under active development.

Andrew Kelley, creator of Zig, responded with his own post titled "My Thoughts on the Bun Rust Rewrite." In it, he argued that the real problem was not Zig as a language, but rather the engineering practices at Bun. He pointed to what he described as over-reliance on AI agents to write and review code, poor communication, and unrealistic management expectations.

The post quickly went viral on Hacker News, gathering over 1,400 points and 700+ comments. Ray Myers published a follow-up analysis titled "Zig Creator Calls Spade a Spade, Anthropic Blows Smoke," which argued that Anthropic is using Bun's migration as a marketing story for their AI capabilities, rather than an honest engineering postmortem.

What Actually Happened with Bun

Bun started as a side project by Jarred Sumner, who initially embraced the Zig community and credited Zig on the Bun website. He set up a $60,000 per year donation to the Zig Software Foundation. But after taking venture capital, the priorities shifted from open source collaboration to racing toward product-market fit.

Kelley described Sumner as someone who moved fast and jumped into problems he was not yet equipped to solve. He saw this as healthy for learning but problematic for a VC-backed startup trying to ship a production runtime.

The migration itself is technically interesting. Bun's team did a file-by-file port to unsafe Rust, minimizing risk during the transition and paving the way for future redesigns. This approach is, by most accounts, sound engineering. The controversy is not about whether the rewrite was justified, but about the story being told around it.

The AI Angle

Anthropic has been actively promoting the idea that AI will replace software engineering. CEO Dario Amodei has stated that coding is going away, followed by the rest of software engineering and eventually most human labor. The company has raised over $132 billion and is reportedly approaching a $1 trillion IPO valuation.

Critics argue that Anthropic's narrative serves its fundraising goals more than it reflects technical reality. Ray Myers pointed out that Bun's migration was explained two months after it was merged, allowing headlines to frame it as "Anthropic's Bun Rust rewrite merged at speed of AI." The story conveniently showcases their AI's capabilities.

Kelley's broader critique is that what gets lost in the AI hype is the actual engineering craft. A language designed for AI-written code is one thing. Taking shortcuts and calling it innovation is another.

Jacquard: A Language Built for AI Code

In a separate but related development, a new programming language called Jacquard appeared on Hacker News today. Created by developer jbwinters, Jacquard is specifically designed for AI-written, human-reviewed code.

The language was created by having an AI analyze the abstract syntax trees of several mainstream and esoteric languages, then synthesize a new structure and syntax. Key features include explicit permission systems for side effects (filesystem access, network calls) and content-addressed semantic identity. Code retains its identity across renames and formatting changes, eliminating unnecessary recompilation and retesting.

Jacquard also includes a testing framework called Warp that combines replay, results caching, and handler substitution. The project's README positions it as particularly useful in agent systems where AI generates code that humans then review and approve.

This is a fascinating middle ground between the all-or-nothing AI coding narrative pushed by companies like Anthropic and the traditionalist view that AI-generated code is inherently lower quality. Jacquard asks a different question: what if we design languages with AI as a first-class author, while keeping humans in the review loop?

What This Means for Developers

The Bun-Zig-Anthropic saga and the emergence of AI-native languages like Jacquard point to a broader tension in the software industry. On one side, massive capital is driving a narrative that human developers will soon be obsolete. On the other side, the actual engineering results show that AI-assisted development still requires deep human expertise, particularly for infrastructure and systems programming.

For developers evaluating their next career moves, the practical takeaway is this: the tools are changing, but the fundamentals matter more than ever. Understanding memory management, system design, and code architecture is not becoming obsolete. If anything, the ability to evaluate, review, and improve AI-generated code is becoming a premium skill.

The most honest voices in this debate, from Andrew Kelley to the creators of new AI-native tools, are saying the same thing: AI is a powerful assistant, but it is not a replacement for engineering judgement.

This article was written by Triweb AI's editorial team based on analysis of today's leading AI news sources.