A neural network is a stack of layers. Raw data goes in the top, each layer transforms it slightly, and a prediction comes out the bottom. The order matters — an output layer before the input would be like reading a book back-to-front. Modern AI models like GPT have hundreds of these layers stacked together.