🧠 Meta-Learning Dynamic Agent Weights

The RIOS feedback loop automatically recalculates Brier error scores after every race. Agents producing real predictive edge receive boosted weights, while noise is suppressed.