LLMs work better in codebases with less tech debt.
Especially as a codebase grows, LLMs can get lost in dead code and red herrings, so I don’t think the fundamental assumption of this article holds true, atleast not in the size of codebases I’m working with.
Especially as a codebase grows, LLMs can get lost in dead code and red herrings, so I don’t think the fundamental assumption of this article holds true, atleast not in the size of codebases I’m working with.