I agree on that, but it's not the only use case. I understand my codebase, I consult with AI on architecture, possible directions to go, upsides and downsides, let it propose a solution, think about it, and then write parts of the code myself and let it write some of the boilerplate which I know exactly how I want it to be. It never does any single task, where I didn't give it precise details and scope what to do and how.
Is the result of this.. slop?