The construction of artificial networks of transcriptional control elements in living cells represents a new frontier for biological engineering. However, biological circuit engineers will have to confront their inability to predict the precise behavior of even the most simple synthetic networks, a serious shortcoming and challenge for the design and construction of more sophisticated genetic circuitry in the future. We propose a combined rational and evolutionary design strategy for constructing genetic regulatory circuits, an approach that allows the engineer to fine-tune the biochemical parameters of the networks experimentally in vivo. By applying directed evolution to genes comprising a simple genetic circuit, we demonstrate that a nonfunctional circuit containing improperly matched components can evolve rapidly into a functional one. In the process, we generated a library of genetic devices with a range of behaviors that can be used to construct more complex circuits.