On December 19, 2018, Federal Reserve Chair Jerome Powell raised rates and described the balance sheet runoff as on "autopilot", signaling that the Fed would continue tightening regardless of market conditions. The S&P 500 sold off sharply, culminating in the worst Christmas Eve session on record. By December 24, the index had fallen approximately 20% from its September high.

Most text analysis tools suffer from what the author calls the “tyranny of topic” — an overreliance on counting words and labeling them as positive or negative. While sentiment analysis can work in tightly constrained contexts (like Federal Reserve minutes or structured product reviews), it collapses under the weight of real-world language.

What does it actually mean to extract insight from words? Most natural language processing tools claim to transform massive volumes of text into actionable data. But in doing so, they often discard the very thing that makes language powerful: meaning. Instead, they measure what’s easiest—topics, sentiment scores, and shallow similarity.