This is made with initial weights to a SIREN style neural network-(ish). Every neuron is connected to every prior neuron. The x,y inputs range from [-1,1]. The three color output is scaled from [-1,1] to [0,255] and clipped. The initial weights take two random integers, divides the smaller one by the larger one, and makes it negative based on which one was smaller. The activation function of each neuron is sin(). This image was done with 100 free neurons. 2 input neurons connect to "free neurons" which also connect to each other, then those free neurons are connected to output neurons.