📄 中文摘要
AI 工具虽然能够提升工作效率，但“差不多正确”的代码可能会在代码审查、测试和代码检查中通过，从而进入代码库，潜伏着引发潜在灾难的风险。尽管行业在推动将控制权交给智能代理，仍需保持审慎和怀疑的态度，信任必须通过实践获得，而不是轻易赋予。建议在使用大型语言模型时，始终保持对生成内容的深度理解和审查，避免因过度依赖而导致的潜在问题。
你的 AI 生成代码“差不多正确”，这实际上比“错误”更糟糕
📄 English Summary
Your AI generated code is "almost right", and that is actually WORSE than it being "wrong".
AI tools can enhance productivity, but code that is 'almost right' can pass reviews, tests, and linters, leading to its inclusion in the codebase and posing risks for potential disasters. While the industry pushes for handing over control to agents, a cautious and skeptical approach is recommended. Trust should be earned through practice and understanding rather than granted. It is essential to maintain a deep understanding and scrutiny when working with large language models to avoid potential issues arising from over-reliance.