Comment by razcle

I agree with that. What do you think about the point thought that for LLM agents and applications, prompts and tool definitions might matter more than code?