Yeah, LLMs are pretty bad at planning maintainable architecture. They don’t refactor it when code is evolving and probably can’t do it due to context limitations.