In the contest, human competitors outperformed AI; most participants even overfitted machine learning models to public data. The investigation showed that all AI agents fell short against the best human efforts.