The Clock and the Pizza: Two Stories in Mechanistic Explanation of Neural Networks