Can we get an AI to write better?

This article explores how to quantify 'style' and 'cadence' in literary works, introducing the Horace project—by analyzing statistical features (entropy, surprisal, cohesion) of texts to identify unique writing fingerprints of different authors. Experiments show that even small models can distinguish Hemingway from Wodehouse using these features. The author believes such quantification can help improve AI writing abilities to produce more stylistically rhythmic text.