Three months ago you built out a project using an LLM. Now every change breaks something, the codebase is full of dead cruft, fallbacks, and partial implementations. Files are 3,000 lines long and everything is a tangled mess. You don't have to learn to code! You don't have to figure this out! You can get out of this mess the same way you got into it, just use the AI.

"The conventional wisdom is that you should slow down, read the agent's output and understand it before committing. That sounds to me like riding a dragon for weeks, and then somebody tells you to get off and lead him around by a leash."