What a neural network actually is — a neuron, a layer, a depth — and how it learns: measuring how wrong it is, sending that error backwards, and stepping downhill without falling over. Built visually, from zero.
