A conservation laws generically develops discontinuities in finite time. For convergence to its weak solution, a numerical method needs to be conservative. A popular way to derive such methods (due to Godunov) is to introduce discontinuities at every cell interface (reconstruction step), and to evolve such step-wise data over a short period of time. Godunov's approach thus introduces discontinuities everywhere in the solution. In view of the big effort associated with grid refinement (particularly in multi-d), efforts are ongoing to guarantee properties of numerical solutions for coarse grids already. It is not surprising that flow phenomena different from shocks (low Mach limit, vortices, ...) are not well approximated by standard Godunov methods on coarse grids. This observation has sparked the development of Active Flux, a numerical method whose degrees of freedom are cell averages and, additionally, point values located at cell interfaces and shared by adjacent cells. The evolution of the averages is conservative, and the method is able to resolve shocks correctly, despite a globally continuous reconstruction. Its centerpiece is a short-time evolution of continuous data. The talk will describe this numerical method, in particular its application to nonlinear conservation laws, as well as recent developments.