📄 中文摘要
许多人使用人工智能助手来编写代码，而作者则选择先让助手故意破坏代码。这一看似反直觉的改变显著降低了错误率，使代码审查变得更有意义。传统的AI编码流程通常包括描述需求、生成代码、审查代码、发现错误和修复错误。然而，AI生成的代码往往在外观上看起来是正确的，导致审查者容易忽视潜在的逻辑错误。作者称这种现象为“正确性剧场”，并强调了调试优先的工作流程在提高代码质量方面的有效性。
优先调试的人工智能工作流程：为何我故意让助手破坏代码
📄 English Summary
The Debug-First AI Workflow: Why I Make My Assistant Break Things on Purpose
Most people use AI assistants to write code, but the author has started using theirs to intentionally break code first. This seemingly counterintuitive change has halved their bug rate and made code reviews more meaningful. The traditional AI coding workflow involves describing requirements, generating code, reviewing it, finding bugs, and fixing them. However, AI-generated code often appears correct, leading reviewers to overlook potential logical errors. The author refers to this phenomenon as 'correctness theater' and emphasizes the effectiveness of a debug-first workflow in enhancing code quality.