>Hardly. Claude Code is basically just a wrapper around an LLM with a CLI.

I don't know why people keep acting like harnesses are all the same but we know they aren't because people have swapped them out with the same models and receive vastly different results in code quality and token use.