📄 中文摘要
在软件开发过程中，最具挑战性的部分并非仅仅是编写代码，而是理解系统、调试复杂问题、设计高效架构以及做出能够避免未来痛苦的决策。大型语言模型（LLMs）虽然可以提供代码建议、帮助生成模板，甚至在某些情况下充当讨论的对象，但它们并不理解系统，也无法在其“思维”中携带上下文，更无法判断决策的正确与否。最重要的是，决策的选择权仍然在开发者手中，真正的软件开发工作依然需要人类的智慧和判断力。
引用大卫·阿布拉姆
发布: 2026年3月23日
📄 English Summary
Quoting David Abram
The most challenging aspects of software development are not merely about typing out code, but rather understanding systems, debugging nonsensical issues, designing robust architectures, and making decisions that can prevent future pain. Large Language Models (LLMs) can suggest code, assist with boilerplate, and sometimes serve as a sounding board, but they do not understand the system, lack context in their 'minds', and certainly cannot discern the rightness or wrongness of decisions. Most importantly, the choice of decisions remains with the developer, emphasizing that the real work of software development still requires human insight and judgment.