Conceptual integrity and counting lines of code

In this article, Simon Willison shares insights from a podcast episode about AI's impact on software development. He argues that lines of code can be a meaningful productivity metric when using coding agents, as they allow engineers to produce significantly more debugged code. However, he emphasizes that the new bottleneck is cognitive capacity, necessitating teams. He also explores the concept of conceptual integrity from 'The Mythical Man-Month', warning that AI agents make it easy to add features rapidly, leading to a 'Winchester Mystery House' effect where software loses coherence and becomes harder to maintain.