The sources and history of cosmic rays are largely unknown because of the intrinsic deflection and delay imposed by cosmic magnetic fields. Minerals might offer a way to investigate this mystery from the rock record, as they preserve the damage tracks of the particles that crossed them over their geological history. This idea, known as paleo-detection, is mostly known for rare-event searches, such as WIMP recoils and neutrino interactions, where nanometre-scale tracks and the need for ultraclean minerals require a demanding readout. Cosmic rays are the opposite case: the signal is expected to be abundant, and the micron-to-tens-of-micron tracks can be revealed with simple etching and optical microscopy.

I will present the simulation framework we built to predict the track-length spectrum for a given mineral and flux history. It combines MCEq and Geant4, follows the overburden history of the sample, and folds in detection efficiency and measurement effects. I will then show how these predictions are confronted with data in PRImuS, our experimental proof of concept: chemical etching of obsidian, argon-plasma etching of halite, automated optical microscopy, and OptimusPrimus, a deep-learning tool for track recognition. The pipeline is validated on obsidian, and its application to halite is the next step. I will conclude with the outlook toward the possible GALACTIC project, based on light-sheet microscopy.