High dark matter fractions make dwarf galaxies ideal laboratories for probing the small-scale structure of dark matter and testing the CDM framework. Shallow gravitational potentials and high gas fractions make dwarf galaxies especially sensitive to the effects of stellar feedback, which can launch large-scale gas outflows from the galaxy into the circumgalactic medium. Sensitivity to feedback is both a blessing and a curse: it means that simulations and observations of dwarf galaxies are ideal for refining feedback models, but it also means that feedback must be understood before observations of dwarf galaxies can robustly distinguish between dark matter theories. I will discuss ongoing efforts to simulate dwarf galaxies, focusing in particular on work from the Feedback in Realistic Environments (FIRE) project. I will pay particular attention to tensions between current simulations and observations, speculating on what is still needed to bring simulations and observations into better agreement.