There is no magic trick

Drawing on the fox-vs-hedgehog framework from 'Expert Political Judgment', the author argues that no single mental model or 'magic trick' reliably predicts outcomes in complex domains like recruiting, investing, or forecasting. After 500+ tech interviews, the author found weak correlation between predicted and actual candidate performance, attributing this to overconfidence and adverse selection in competitive markets. The key insight: collect many independent signals and combine them, mirroring the machine learning principle of ensemble methods and boosting — where aggregating many weak learners consistently outperforms any single model.