From misty origins a business emerged. The company’s main epistemological bet was that the history of social media could be studied to predict the future. Through the application of machine learning, a branch of artificial intelligence, social media signals could be used to predict events—all sorts of events, from protests and strikes to terrorist attacks, election outcomes, and financial market moves—before they materialized in real life. The system’s algorithm was trained to get better at recognizing the characteristic patterns of online activity that precede events, and to alert users whenever those patterns were beginning to recur. This “predictive intelligence” tool, as Jim called it, was then to be packaged and sold to government intelligence and defense agencies, hedge funds and investors, and a host of other deep-pocketed corporate worthies.