Vibe Coding cripples the mind

A viral critique on LessWrong titled 'Vibe Coding cripples the mind' has sparked debate about the long-term impact of AI-assisted programming. The author, 'spookyuser,' recounts helping a hackathon team with a project that 'kind of worked' but was a maze of 50+ uppercase markdown files and a sprawling 3,400-line main script, making deployment a multi-hour ordeal. The post argues that 'vibe coding'—a term coined by Andrej Karpathy in 2025 describing coding where you 'embrace exponentials, and forget that the code even exists'—is creating a generation of programmers who can produce functional outputs but lack the underlying skills to create coherent, maintainable systems. The author's investigation, using a script to find similar GitHub repos, suggests this chaotic style is becoming common.

The analysis draws a parallel to programmers no longer needing to write assembly language, but warns the comparison is flawed because game engines produce optimized, reliable code, whereas current AI coding assistants often generate messy, human-incomprehensible output. The post cites an Anthropic study where students using AI averaged 50% on a quiz versus 67% for hand-coders, indicating a significant knowledge gap. While acknowledging past warnings about new abstractions (like Dijkstra's skepticism of COBOL) were often wrong, the author contends 'vibe coding' is different because it bypasses foundational problem-solving, leading to systems that are impossible to debug or extend, ultimately creating massive technical debt for the industry.