In the watery environment, vortices batter the mechanical fish. The AI agent, like an experimental chef, ran simulations, tried out 'recipes' of movements, and improved the program. Each experiment was like spectroscopy: the agent broke down behavior into elements, uncovering patterns. At first, the fish only paddled in one direction. But after trying hundreds of combinations, the agent derived a universal algorithm for any direction. The result: a clear instruction — body wave, aim, stabilize. This reduced the entropy (disorder) of movements to a minimum.
Astoundingly, the agent invented a swimming method not found in nature: a simultaneous body wave and target tracking, like a homing torpedo. It logged all its 'recipes,' making the AI's logic more transparent than any neural network.
Unlike neural networks, whose inner workings remain hidden, this agent left a detailed diary of all decisions, like an open cookbook.