I think getting the initial domain modeling, structure and opinions that build up a code base is more of an art than science. It’s highly opinionated. Sure you

I think getting the initial domain modeling, structure and opinions that build up a code base is more of an art than science. It’s highly opinionated. Sure you have SOLID but in my experience no one follows it to the book. Once the code base structure and opinions are in place, I think LLMs are decent at writing code that are bounded to a single concern and not cross cutting. LLM generated code bases work initially but so will code written by college kids for an initial set of requirements. Even a staff+ level engineer will struggle in contributing to a code base that is a mess. Things will break randomly. Don’t see how LLMs are any different.