A memory is not a file tucked away in a single cell. Donald Hebb's 1949 rule says it plainly: neurons that fire together, wire together. When a pattern of neurons is active at once, the connections between them strengthen — and the memory now lives in those connections, not in the neurons themselves.
This is a Hopfield network (1982): 64 neurons, every pair joined by a weight. Teaching a pattern adds its Hebbian outer product to the weights. Wipe the neurons and the weights are untouched — so paint back a fragment and the network slides “downhill” in energy into the nearest stored pattern, rebuilding the whole thing. That is why retrieval is reconstruction, not playback.
But every memory shares the same weights. Past a capacity of about 0.14 × N ≈ 9 patterns, recall lands between memories — a blend. The same neurons that carry one memory carry the others, which is exactly why crowded memories bleed together: interference.
Something in the simulation stopped unexpectedly — the lesson continues without it. You can move on; nothing you did was wrong.