In my talk, I argue that in order to solve the problem of electronic correlation in solids, we need to understand and leverage its structure. To that end, I expose the idea of "latent space" and aim to show its connection to the ancient greek theory of Forms, to techniques already used in physics and chemistry, and to modern machine-learning. Finally, I show how the separation and coupling of length and time scales prevalent in physics can be understood as such latent spaces. This will give rise to quantics tensor trains (QTTs), a powerful and promisting tool for tackling the problem of many correlated electrons.