Management thinks a crutch can effectively replace people massively in sensitive knowledge work. When that crutch starts making errors that cost those businesse

Management thinks a crutch can effectively replace people massively in sensitive knowledge work. When that crutch starts making errors that cost those businesses millions, or billions, well, hopefully management who implemented all that will get fired... Yes, LLM are useful, but they are even less trustworthy than real humans, and one needs actual people to verify their output, so when agents write 100K lines of code, they'll make mistakes, extremely subtle ones, and not the kind of mistake any human operator would make.