📄 中文摘要
当前的 AI 编程工具主要依赖源代码进行操作，能够读取代码、预测应用功能并生成编辑。这在处理纯逻辑时表现良好，例如具有明确输入和输出的函数、重构和类型级别的变更。然而，当源代码无法完整描述应用程序的运行时环境时，这种方法就会失效。对于任何具有运行时的应用程序，如在浏览器中运行的 web 应用、处理请求的服务器以及具有中间件和编译输出的框架，源代码从来不是完整的故事。前端的问题显而易见，浏览器中的计算样式和渲染的 DOM 并不存在于源文件中。服务器端同样存在其独特的运行时上下文，源代码无法捕捉到注册的路由、编译后的模块图以及其他关键信息。
📄 English Summary
AI Coding Tools and the Runtime Context Gap
Current AI coding tools primarily operate on source files, enabling them to read code, predict application behavior, and generate edits effectively for pure logic scenarios, such as functions with clear inputs and outputs, refactoring, and type-level changes. However, this approach fails when the source code does not encapsulate the entire runtime context of an application. For any application with a runtime—such as a web app running in a browser, a server handling requests, or a framework with middleware and compiled output—the source code is never the complete picture. This issue is not limited to the frontend, where computed styles and rendered DOM elements are absent from source files. The server side also has its own runtime context that source code alone cannot capture, including registered routes, the structure of the compiled module graph, and other critical elements.