When the human mind is free to roam, its subjective experience is characterized by a continuously evolving stream of thought. Although there is a technique that captures people’s streams of free thought—free association—its utility for scientific research is undermined by two open questions: (a) How can streams of thought be quantified? (b) Do such streams predict psychological phenomena? We resolve the first issue—quantification—by presenting a new metric, “forward flow,” that uses latent semantic analysis to capture the semantic evolution of thoughts over time (i.e., how much present thoughts diverge from past thoughts). We resolve the second issue—prediction—by examining whether forward flow predicts creativity in the lab and the real world. Our studies reveal that forward flow predicts creativity in college students (Study 1) and a representative sample of Americans (Study 2), even when controlling for intelligence. Studies also reveal that membership in real-world creative groups—performance majors (Study 3), professional actors (Study 4) and entrepreneurs (Study 5)—is predicted by forward flow, even when controlling for performance on divergent thinking tasks. Study 6 reveals that forward flow in celebrities’ social media posts (i.e., on Twitter) predicts their creative achievement. In addition to creativity, forward flow may also help predict mental illness, emotional experience, leadership ability, adaptability, neural dynamics, group productivity, and cultural success. We present open-access online tools for assessing and visualizing forward flow for both illustrative and large-scale data analytic purposes.