The accuracy of the interpretation of the current generation of late-time cosmology probes is greatly limited by the theorists' ability to predict the response of baryons. This will only get more difficult in the coming years with the upcoming surveys (DESI, Euclid, LSST, 4-most) mapping our sky with unprecedented precision. In this talk, I will discuss some of the efforts towards this challenge on the simulation side. I will, in particular, focus on the latest generation of large-scale hydrodynamical simulations, the FLAMINGO suite, designed to be virtual twins of the Stage IV cosmology surveys. This suite includes the largest simulation ever run to z=0 and exploits a state-of-the-art baryon physics model calibrated to the relevant data.Variations in the physics, trained via machine learning techniques, allow us to encompass the uncertainty in the modeling. Outputs in observer space, such as X-ray, stellar light and SZ lightcones are produced and are used to validate the simulations. I will present selection of key results obtained in the last year or so by exploiting the simulation data to help answer some of the current puzzles in late-time cosmology.
This talk is also a invitation to make use of this data for your research.