Stop generating, start thinking! We do not need more AI code, we need more human thinking about the code we ship. Spicy autocomplete is fine but outsourcing our judgment is not. Especially since LMM are trained on bad code to start with. Do we really want to have them repeat our human mistakes?
