Came back to scratch and made this for fun in a few hours. A super simple supervised learning model trains to create the best path. GREEN FLAG to start project SPACE BAR toggles training TO DISPLAY SLIDER shows all of the AI models mid training. 1 is the best working model, 50 is the worst. 51-100 are mutations of 1-50 F shows frame rate/tickrate (I think those two line up in scratch correct me if I'm wrong) Turbo Mode recommended (Shift+Green flag) UPDATE: I've been using this model to test different reward functions. I've been trying to find the most consistent reward function myself, but I've also been testing different LLMs to see if they can come up with a good one.
