一名身处网络封锁环境的伊朗开发者在开发离线Zig语言音乐播放器AMP时,过度依赖Anthropic的Claude Max 20x订阅进行“氛围编码”(vibecoding),导致生成的代码库难以维护 1。该项目使用Zig编写,包括着色器部分 1。然而,由于开发者居住在伊朗,Anthropic突然封禁了其账户 1。与此同时,代码托管平台Codeberg宣布实施无AI政策,进一步迫使该开发者移除所有由AI生成的代码 1。
面对双重困境,开发者不得不手动重写项目,历时四个月最终删除了近一半的代码库 1。尽管经历了繁琐的重构过程,但清理后的程序展现出更强的鲁棒性 1。这一案例深刻揭示了盲目使用大型语言模型(LLM)进行软件开发的风险,强调应将AI视为辅助性的IDE功能,而非自动化的程序员替代品 1。
An Iranian developer encountered significant technical and administrative challenges while developing AMP, an offline music player written in the Zig programming language, which includes shaders 1. During a prolonged period of internet restrictions in Iran, the developer utilized an Anthropic Claude Max 20x subscription to assist with coding, resulting in a codebase heavily generated by AI through a practice known as "vibecoding" 1. This reliance on automated generation ultimately made the project difficult to maintain 1.
The situation escalated when Anthropic immediately banned the developer’s account due to their residence in Iran 1. Compounding these issues, the hosting platform Codeberg announced a policy prohibiting AI-generated code, forcing the developer to remove all such content from the repository 1. Consequently, the developer was required to manually rewrite and clean up nearly half of the codebase over a four-month period 1. Although this process involved deleting approximately 50% of the original code, the final application emerged more robust than before 1. The case highlights the risks of blindly using large language models for software development and emphasizes the importance of treating AI as a supplementary IDE feature rather than an autonomous programmer 1.
评论
还没有评论,欢迎留下第一条。