Synthetic biology holds the promise of allowing us to engineer living beings. I will start by reviewing some examples where mathematical models lead to the development of synthetic organisms with particular properties: One such example is a synthetic gene oscillator in Escherichia coli that exhibits robust temperature compensation -- it maintains a constant period over a range of ambient temperatures. A mathematical model predicted and experiments confirmed the particular mechanisms that lead to temperature compensation despite Arrhenius scaling of the biochemical reaction rates. Such successes are encouraging. But how far can our theoretical models take us? I will argue that our models are still fairly coarse, and do not adequately describe all the important properties of genetic signaling networks. For instance, "transcriptional delay" - the delay between the start of protein production and the time a mature protein finds a downstream target - can have a significant impact on the dynamics of gene circuits. Such delay can inhibit transitions between states of bistable genetic networks, as well as destabilize steady states in other networks. I will show how these effects can be described by reduced, non-Markovian models that are quite different from established models. I will also discuss work with experimental collaborators to characterize the distribution of this delay.