There is ongoing work worldwide to write new dynamical cores for numerical weather prediction. The reasons are simple refactoring, take into account new processor architectures, try new programming environments, and finally use latest achievements in numerical mathematics. I will summarize actual developments and show some examples from my own personal endeavor within the Climate Modeling Alliance (https://clima.caltech.edu/). Here we develop a new numerical core using the programming language Julia for a new earth system model which should learn from different data sources. The new dynamical core is based on a cubed sphere grid with high order continuous or discontinuous ansatz functions. To understand stability issues I have implemented standard test cases in the DG code FLUXO. Here the same stability issues were observed but could be resolved with a so called split-form kinetic energy conserving formulation. For a second planned formulation with continuous elements I have implemented my algorithm version in Matlab and subsequently in Julia. By means of the Held-Suarez example we will compare implementation details to get efficiency for both programming environments and present a new Rosenbrock-W-method where the explicit part has the strong stability preservation (SSP) property.