Every child learns early that brittle objects break easily. Yet on a bit of reflection the process should seem mysterious. Weak forces are applied over large length scales to an object, and spontaneously the object uses that energy to snap atomic bonds. It happens through propagation of a crack, a propagating structure with a singularity at its tip. George Irwin created an ingenious theory for how cracks propagate, posed entirely in the framework of continuum mechanics, which cleverly evades all questions about what happens very near the singular tip of the crack. However, there have long been some dynamical puzzles about how cracks moved. For example, they seemed to move at half the speed that theory predicts. I will argue that these puzzles are the result of trying to push continuum mechanics too hard, and will present analytical, numerical, and experimental evidence that one can understand crack motion rather thoroughly when continuum mechanics is replaced by the more realistic underlying discrete mechanics.