Despite the substantial progress on the Minimal Model Program for three-dimensional varieties in characteristic p>5, so far little was known for p <= 5. The goal of this talk is to shed some light on the geometry of threefolds in low characteristic. This is based on joint work with Christopher Hacon.