A Human in Control

> Not to mention how I want to know how curl works, even if I don’t keep 100% intimate knowledge of every single angle and corner, I know most of it. I think it helps me make better decisions, debug better, help users better and keep the architecture sound.

> Getting the initial code written is not the big deal. For curl, maintaining and polishing the landed code through decades is the real task.

This is the real sticking point for me with vibecoding. In the short term, vibecoding everything might get you something that works. But I have a hard time accepting that in the long-term, maintaining a large and complex codebase is feasible when you haven't actually developed a deep understanding of it.

But maybe that's my bias as someone who considers their job description to be basically "develop a deep understanding of and maintain this large codebase".