How to keep AI-written code aligned (without repeating yourself)

I’m the author. This came out of watching AI tools subtly rewrite system boundaries while still passing tests. The post is about documentation topology and keeping intent close to code—not about a specific tool.