This study presents an innovative educational model that integrates Artificial Intelligence (AI) into the teaching of differential equations through experimental validation in a school wind tunnel. Conducted with three student groups from the MA1033 course at Tecnológico de Monterrey (Campus Puebla), the project explored the aerodynamic drag and lift governed by nonlinear differential equations derived from Newton's Second Law and Bernoulli's principle. Each group interacted with AI at a different cognitive level: analytical regression, predictive modeling, and explainable deep learning. The results demonstrated that the progressive integration of AI significantly improved both modeling accuracy (average R2> 0.97) and cognitive engagement. Beyond data processing, AI acted as a cognitive mediator that enabled students to interpret residuals, identify anomalies, and reason about the limitations of theoretical models. The combination of experimental observation, mathematical modeling, and algorithmic interpretation fostered the development of adaptive reasoning, data ethics, and scientific communication skills. This hybrid AI-experimental learning framework redefines engineering education by transforming AI from a computational tool into an epistemic collaborator, enhancing conceptual understanding, reflective thinking, and metacognitive awareness in scientific learning.
