Public talk aiming to explain how cosmological constraints are obtained. After motivating that different cosmological parameters influence the visual appeareance of dark matter snapshots from a simulation, we show that this visual difference can be compressed into summary statistics like the 3D correlation fuction. We then show the cosmology-dependence of the correlation function and use it to visualize the likelihood contours in the Omegam-S8 plane. We then present an interpret the cosmological constraints of the KiDS-1000 analysis and give an outlook about the prospect of future missions, such as Euclid.