Active Flux and Discontinuous Galerkin look like different methods. Active Flux uses globally continuous approximations with point values at cell interfaces. Discontinuous Galerkin allows discontinuities between cells and uses polynomial bases within each cell. Different philosophies, different implementations, different communities.
Note (2026-11-12; read against the paper's full text): the equivalence is for the semi-discrete versions of Active Flux and Discontinuous Galerkin (the method of lines with Runge-Kutta time integration). The paper says the original Active Flux method uses evolution operators and is a single-stage fully discrete method, which is not what it shows to be equivalent. By equivalence it means that for a given DG method and a given PDE there exists an Active Flux-type method that agrees with it, and it states the result for linear and nonlinear systems in one dimension and for linear advection on Cartesian meshes in two dimensions; it notes that the simplest form of the result, for linear advection in one dimension, already appears in earlier work by Roe. The essay's statement that the authors prove the two methods are the same, without these conditions, goes beyond the paper; the paper also reports numerical experiments on the cost-error trade-off.
Barsukow, Klingenberg, and Krotsch prove they are the same method.
For linear problems in one and multiple dimensions, a mapping exists between the degrees of freedom of Active Flux and DG such that the updates agree exactly. The two methods, derived independently from different principles, produce identical numerical solutions. The mapping extends, in some sense, to nonlinear problems.
The equivalence reveals structure that neither method shows alone. Radau polynomials and their zeros appear naturally in the AF-DG mapping — objects well-known in DG theory but with no obvious role in Active Flux. Their appearance explains DG superconvergence: the phenomenon where DG solutions converge faster at certain points than the global rate would suggest. Active Flux “functions as the underlying high-order method within DG” — the superconvergence points are where the Active Flux structure shows through the DG formulation.
Active Flux is also more economical. With K+1 degrees of freedom per cell on average, it achieves polynomial degree K+1, while DG uses the same number for degree K. The continuous interface values in Active Flux are shared between cells; DG's discontinuous formulation doubles them. Same computation, higher approximation order.
Two communities building separate methods were building the same method from different directions. The equivalence was always there, hidden by the difference in formulation.