Large language models can help a lot, yet they still lack the human touch, particularly in the areas of context comprehension and question formulation. The entire "no copy-paste" rule seems strange as well. It is as if the models were performing an operation solely in their minds rather than just repeating it like we do. It gives the impression that they are learning by making mistakes rather than thinking things through. They are certainly not developers' replacements at this point!