> It’s just code

You're not informed about what LLMs are. LLMs are not "code" in the imperative sense (although "code" is used run them), and that's a big problem - since we never had neural networks of such scale, there is confusion about categorizing them and most importantly, their behavior.