📄 中文摘要
该研究提出了三个相互关联的假设，探讨可执行规范在 AI 辅助软件开发中的作用。第一个假设是相关错误假设，强调了在代码审查过程中，错误之间的关联性如何影响软件质量。第二个假设基于复杂性科学，指出在复杂系统中，规范的执行能够有效降低错误发生的概率。第三个假设则映射了规范无法捕捉的残余缺陷，强调了在软件开发中，尽管有规范的指导，仍然可能存在未被发现的缺陷。这些假设共同构成了对 AI 辅助代码审查的完整论证，旨在提升软件开发的质量和效率。
规范作为质量门：关于 AI 辅助代码审查的三个假设
📄 English Summary
The Specification as Quality Gate: Three Hypotheses on AI-Assisted Code Review
This research develops three interconnected hypotheses regarding the role of executable specifications in AI-assisted software development. The first hypothesis, the correlated error hypothesis, emphasizes how the interrelation of errors during code review impacts software quality. The second hypothesis is grounded in complexity science, indicating that the execution of specifications can effectively reduce the likelihood of errors in complex systems. The third hypothesis maps the residual defects that specifications cannot catch, highlighting that despite the guidance of specifications, undetected defects may still exist in software development. Together, these hypotheses form a comprehensive argument for AI-assisted code review, aiming to enhance the quality and efficiency of software development.