First Tragedy, then Parse: History Repeats Itself in the New Era of Large Language Models