This mini-course is concerned with those PDEs which have a gradient flow structure in the Wasserstein space W_2 and can thus be attacked via the so-called Jordan-Kinerlehrer-Otto scheme, a sequence of iterated minimization problems in the space of measures which provide a time-discretization of the solution. The first lecture will explain the convergence of this scheme to a solution of the PDE and present the main techniques to obtain it, including in the case where the distance W_2 is replaced with W_p, which lets non-linear PDEs of p-Laplace type appear. The second lecture will introduce some tools to prove iterable estimates on the solution of the JKO scheme. We will then show how one can easily recover well-known results for linear diffusion but also new estimates which would be more difficult to guess using "continuous" methods rather than this time-discrete counterpart.