> I'd wander into my office, check what Claude had built, test it real quick. If it worked, great! Commit and push. Do people read the code? Or just test if it

> I'd wander into my office, check what Claude had built, test it real quick. If it worked, great! Commit and push. Do people read the code? Or just test if it work and push? To me, code is like a map that has to be clear enough so other humans can read it to navigate the territory (codebase). Even if it's just two – me and AI agent – working on the codebase, it's not much different from "me and another programmer". We both want to have updated mental model of how exactly code structured and how it works and why. Using AI for coding and not reading the code sounds more like stopping being developer and self-promoting yourself to the manager of AI-programmers who trusts their craft completely.