In this paper, an optimization-based approach to construct spiking networks for the purposes of decoding and control is presented. Specifically, we postulate a simple objective function wherein a network of interacting, primitive spiking units is decoded in order to drive a linear system along a prescribed trajectory. The units are assumed to spike only if doing so will decrease a specified objective function. The optimization gives rise to an emergent network of neurons with diffusive dynamics and a threshold-based spiking rule that bears resemblance to the Integrate and Fire neural model.