📄 中文摘要
在 AI 提交的代码中，虽然代码整洁且符合提示要求，所有单元测试也通过，但却使用了团队上个季度已弃用的库，并且违反了服务架构文档中的设计模式。这种情况揭示了提示工程的局限性，强调了上下文工程的重要性。AI 工具在生成代码时，常常需要进行仔细的手动修正，反映出在特定系统上下文中，代码的正确性并不仅仅依赖于代码本身的质量，而是与系统的整体架构和设计规范密切相关。
上下文工程与提示工程：构建 AI 系统的转变
📄 English Summary
Context Engineering vs Prompt Engineering: the shift in how we build AI systems.
The article highlights the limitations of prompt engineering in AI code submissions. Although the AI-generated code appears clean, adheres to prompts, and passes all unit tests, it may still utilize deprecated libraries and violate architectural design patterns. This situation underscores the importance of context engineering, suggesting that the correctness of AI-generated code is not solely dependent on its quality but also on its alignment with the overall system architecture and design specifications. The ongoing debate between context engineering and prompt engineering is rooted in the practical challenges faced when integrating AI tools into existing systems.