Kinda mixed feelings about using AI. For writing it’s definitely bad. But I think it’s totally fine to use an LLM to gauge which of two (anonymized) FOMC speeches is more hawkish (macro context dependent) and repeat that thousands of times to derive an ELO-like score for each.(Joe Weisenthal)