This study systematically evaluates and enhances the performance of the Harris Hawks Optimization (HHO) algorithm across twentythree benchmark functions. The primary objective is to minimize solution errors and optimize convergence by repeatedly testing functions F1 to F23 in MATLAB, incorporating AI-driven code improvements. These enhancements focus on dynamic parameter adjustments and escape mechanisms to avoid local optima, effectively mimicking HHO’s cooperative hunting strategy. Preliminary results reveal a 15–22% decrease in fitness values compared to the basic HHO, particularly when applied to multimodal functions such as F7 and F15. This approach demonstrates the efficacy of iterative testing and machine-learning-based code optimizations in developing advanced metaheuristic algorithms for real-world optimization challenges