I’m very excited about our new preprint, Discretize first, filter next: learning divergence-consistent closure models for large-eddy simulation. We propose a new paradigm, “discretize first, filter next” to learn closure models for large eddy simulation that are fully model-data consistent. It allows for employing neural closure models in the same environment as where they have been trained. We shot that our neural closure formulation produces stable and accurate results for both a-priori and a-posteriori training, without requiring stabilization techniques.