All plants and animals and many unicellular organisms possess circadian clocks-autonomous oscillators with a roughly 24 hour period that allow them to anticipate daily cycles of light and dark. I will discuss recent progress on understanding one such biological clock, in the photosynthetic bacterium S. elongatus. This system has the remarkable feature that the core biochemical oscillator can be reconstituted in vitro with only three purified proteins. Thus, unlike almost all other circadian clocks studied to date, it requires neither transcription nor translation but functions entirely post-translationally. After reviewing what we know about how the in vitro oscillator functions, as well as a few outstanding puzzles, I will turn my attention to the implications of this understanding for clock function in the living cell. In particular, I will argue that the core post-translational oscillator is necessary to make the clock robust to several perturbations present in any growing, dividing cell, but that other specific adaptations are also required. These include negative feedback on the transcription of clock components and the presence of several identical copies of the bacterial chromosome in the same cell. I will conclude by suggesting that, far from being an isolated case, the workings of the circadian clock built entirely on protein modifications actually point the way towards a new understanding of a number of other oscillatory biological systems, which can all be thought of as examples of a new class of "molecular synchronization oscillators."