The creator of the Zig programming language has lambasted the developers of Bun, a popular JavaScript runtime, after they used Anthropic's Claude AI model to rewrite their codebase from Zig to Rust in just 11 days — at an API cost of roughly $165,000. Andrew Kelley, who created Zig, described the resulting code as 'unreviewed slop', arguing that relying on an AI model to produce a complete port without human oversight risks introducing subtle bugs and long-term maintenance headaches.
Bun, originally built in Zig to maximise performance, is a runtime designed to speed up JavaScript and TypeScript development. The team behind it decided to port the core to Rust, using Claude to generate the bulk of the new code. While the speed and cost of the rewrite are notable — $165,000 at API pricing is a fraction of what a human engineering team would charge — critics warn that AI-generated code lacks the rigorous review process needed for production-grade software, especially in performance-critical tools like Bun.
For UK businesses, the controversy underscores a broader tension in the tech sector: the allure of rapid, cost-effective AI-assisted development versus the risks of deploying code that hasn't been thoroughly vetted by human experts. 'This is a cautionary tale for CTOs and engineering leads,' said Dr. Eleanor Shaw, a software engineering researcher at the University of Cambridge. 'AI can accelerate prototyping, but treating its output as a finished product without proper review is a gamble that could lead to security vulnerabilities or reliability issues down the line.'
The UK's Information Commissioner's Office (ICO) has yet to issue specific guidance on AI-generated code, but the incident may prompt renewed scrutiny. Under the EU AI Act, which also affects UK firms operating in Europe, high-risk AI systems — including those used in critical software development — must meet transparency and human oversight requirements. While the Act does not directly cover AI-assisted coding tools, regulators are watching how companies use generative AI in production environments.
For UK consumers, the immediate impact is minimal, but the long-term implications are significant. If AI-generated code becomes commonplace in everyday applications — from banking apps to smart home devices — the risk of undetected flaws could erode trust in digital services. Conversely, if managed properly, AI could lower software development costs, potentially leading to cheaper and faster innovation for British startups and enterprises alike.
Kelley's critique has reignited a debate within the developer community about the limits of AI in software engineering. While some argue that tools like Claude can boost productivity, others insist that human judgement remains irreplaceable, particularly in complex systems where correctness is paramount. For now, the Bun project continues, but the controversy serves as a stark reminder that AI-generated code is only as reliable as the review process behind it.