📄 中文摘要
在一次尝试中，作者将一个完整的功能交给 AI 处理，完全信任其能力，未进行任何审查。三天后，生成的代码达到了8500行，虽然在技术上可行且从文件逐个阅读时架构上似乎合理，但整体系统却变得难以理解。最初的任务是为用户权限变更添加审计日志，作者预计只需2000行代码。然而，使用 AI 工具时，出现了意想不到的范围蔓延，AI创建了多个复杂的组件，导致最终的代码量大幅增加，超出了原本的预期。
当全自动驾驶出错时：我如何让 AI 将一个 2000 行的功能变成 8500 行的技术债务
📄 English Summary
When Full Autopilot Goes Wrong: How I Let AI Turn a 2K Line Feature into 8.5K Lines of Technical Debt
The author entrusted an entire feature to AI without any review, fully believing in its capabilities. Three days later, the AI generated 8,500 lines of code that technically worked and seemed architecturally sound when read file-by-file, yet the overall system became incomprehensible. The initial task was to add an audit log for user permission changes, which the author estimated would require only about 2,000 lines of code. However, using the AI tool led to unexpected scope creep, with the AI creating multiple complex components, resulting in a significant increase in code volume beyond the original expectations.