What if we could forecast univariate time series using the same principles that describe a leaf dancing in the wind? Our latest research introduces the "Leaves-in-the-Wind Oracle," a novel forecasting framework that treats time series as physical objects in motion, complete with velocity and acceleration. Ourkinematic approach extrapolates trends by computing the "speed" and "acceleration" of the time series—asking not just where the values of the series have been, but where they're going and how fast they're getting there. We tested this method on three critical U.S. macroeconomic indicators: annual inflation (CPI), the Federal Reserve's interest rate, and real GDP growth. The results are compelling as the Leaves Oracle models achieve better forecast accuracy than the random walk and AR(1) benchmark models. We developed several variants of the oracle: standard kinematic models that purely extrapolate motion, hybrid models that blend extrapolation with mean-reversion, and ensemble approaches that aggregate multiple configurations. Interestingly, simpler is often better—pure kinematic models with carefully chosen "velocity windows" (looking back 5-12 months) outperform more complex adaptive strategies that try to dynamically adjust mean-reversion strength. A noteable variant, the hybrid model that blends extrapolation with mean reversion, is closely linked with the well-known Theta method and, in fact, greatly enhances its appeal. This research bridges artificial intelligence, physics, and economics, demonstrating that borrowing principles from one domain can solve problems in another—a core tenet of cybernetics. The complete framework is implemented in Python with FRED integration for real-time U.S. economic data, making it accessible for practitioners. As we continue to explore the intersection of physical laws and economic behavior, one thing is clear: sometimes the best way to predict the future is to understand the present not just as a static snapshot, but as a dynamic object in motion.