This document proposes automating the evolution of neural networks through genetic algorithms. It discusses defining nodes and clusters, connecting them with weights that could operate according to quantum principles, and using genetic algorithms to generate new network architectures by combining codes from previous generations. The goal is to evolve increasingly intelligent networks, where problems could be generated by other networks through adversarial training without human input. Key challenges include defining what makes a network "intellectual" and appropriate nodes and architectures for automated evolution. Coding frameworks like Chainer and TensorFlow could support implementing and testing this approach.