Fracture propagation in brittle materials has been studied using hybrid molecular dynamics simulations, allowing the use of quantum calculations wherever and whenever needed by the problem under investigation. Several timescales of the fracture process have been taken into consideration, going from the lowest possible velocity for a crack growing in subcritical loading conditions, to the highest velocity computed for very cold cracks. The study has been carried out in two prototypical models of brittle fracture: silicon and amorphous silica. Aiming to interpret the lowest crack velocity ever measured by experiments on silicon at room conditions, we used quantum-accurate computer simulations to show that immediate dissociation of oxygen molecules, and consequent oxidation of the highly stressed silicon crack tips, may be the cause of the observed slow crack growth. This theoretical prediction, supported by experimental evidence, claries longstanding discrepancies concerning the role of oxygen as a stress corrosion agent in silicon. Turning the attention to fast crack behaviour, a crossover between activated and catastrophic branches of crack velocities as a function of temperature has been detected in a hybrid classical molecular dynamics model of silicon. Cold cracks travel faster for high loading energies, while this trend is reversed in the region of energies where activated processes become dominant. Finally, the study of catastrophic fracture in amorphous silica has been initiated, testing for the rst time our hybrid approach to non crystalline structures.