I've yet to hear an LLM-generated podcast about an article that wouldn't have been better if the AI had simply read the article aloud.