The equilibrium Ising model has applications to ferromagnetism, lattice gases, neuroscience, etc.; most importantly, it shaped our understanding of phase transitions. In this talk I'll discuss its non-equilibrium version, specifically Ising ferromagnets quenched to zero temperature. It was considered obvious that a ground state is necessarily reached. Surprisingly, in two dimensions other outcomes (stripes) arise with a finite probability; in three dimensions a ground state is never reached for an infinite system. The situation in two dimensions is better understood thanks to a tantalizing connection with equilibrium critical percolation. A general lesson that the gradient descent often cannot find the minima of the cost function and that the algorithm can get trapped and wander indefinitely in a set of the same cost states goes beyond Ising systems.