Yeah, I get this too. Still, I think sometimes being forced to grind on something will spur the "oh wait" moment that leads to new ways of thinking about things. Whereas when the LLM is doing the grinding, you don't see it. You just get a final PR with only the answer to the problem at hand, and you miss the bigger opportunity.

That said, maybe it's not a big deal. Kind of like way back when I moved from C++ to GC code, I remember I missed memory leaks, because having it all automatically taken care of for free felt like giving up control and encouraging of lazy practices and loose ends. Turns out it wasn't really a big deal at all.