## I Trained a Tiny Network to Compress Data. It Drew a Pentagon. Reproducing Anthropic's "Toy Models of Superposition" from scratch in NumPy, with hand-derived gradients and no borrowed numbers. I wasn't expecting geometry to show up. I was reproducing a small piece of Anthropic's 2022 interpretability paper, "Toy Models of Superposition" [1], mostly because the central claim sounded implausible enough that I wanted to check it myself rather than take it on faith. The claim: a neural network can represent more features than it has dimensions to work with, by packing them in at angles to each…