Not to mention these patterns didn't come out of thin air. The LLMs are statistically regurgitating language from its training set, which researchers probably tuned more towards journalistic sources like the one we're reading.