📄 中文摘要
使用大型语言模型（LLMs）生成代码时，通常会面临一种“老虎机”工作流程。通过提示拉动杠杆，可能获得良好的结果，但在不同的模型或与不同的同事合作时，相同的请求却可能产生截然不同的结果。这种不一致性在软件工程中被称为“模糊税”。根本原因在于将意图与实现混淆，自然语言本身具有固有的模糊性。例如，常见请求“实现一个带验证的用户资料页面”会留下许多关键问题未解答，导致AI生成的代码缺乏一致性和可预测性。
第一部分：意图工程系列 - 停止提示，开始编译：可预测的 AI 生成代码之路
📄 English Summary
Part 1 of 3 — Engineering Intent Series - Stop Prompting, Start Compiling: The Path to Predictable AI-Generated Code
When using large language models (LLMs) to generate code, users often encounter a 'slot machine' workflow. Pulling the lever with a prompt may yield a great result, but the same request can produce entirely different outcomes when using a different model or collaborating with different colleagues. This inconsistency is known as 'The Ambiguity Tax' in software engineering. The root cause lies in conflating intent with implementation, as natural language is inherently ambiguous. For instance, a common request like 'Implement a user profile page with validation' leaves numerous critical questions unanswered, leading to a lack of consistency and predictability in AI-generated code.