LLMs and programming languages; interpretability