> LLMs are still not good enough where you can unleash it upon a huge codebase and get good results btw.

Let’s just say that both OpenAI and Anthropic strongly disagree with this, in theory and in practice.