The variational inequality problem can be formulated as a nonconvex, nondifferentiable optimization problem. We show that this formulation leads to a convergent algorithm that decreases the "gap" function in a monotone fashion.