At a very basic level, physics is about what we can say about propositions like \'A has a value in S\' (or \'A is in S\' for short), where A is some physical quantity like energy, position, momentum etc. of a physical system, and S is some subset of the real line. In classical physics, given a state of the system, every proposition of the form \'A is in S\' is either true or false, and thus classical physics is realist in the sense that there is a \'way things are\'. In contrast to that, quantum theory only delivers a probability of \'A is in S\' being true. The usual instrumentalist interpretation of the formalism leading to these probabilities involves an external observer, measurements etc.In a future theory of quantum gravity/cosmology, we will have to treat the whole universe as a quantum system, which renders instrumentalism meaningless, since there is no external observer. Moreover, space-time presumably does not have a smooth continuum structure at small scales, and possibly physical quantities will take their values in some other mathematical structure than the real numbers, which are the \'mathematical continuum\'. In my talk, I will show how the use of topos theory, which is a branch of category theory, may help to formulate physical theories in a way that (a) is neo-realist in the sense that all propositions \'A is in S\' do have truth values and (b) does not depend fundamentally on the continuum in the form of the real numbers. After introducing topoi and their internal logic, I will identify suitable topoi for classical and quantum physics and show which structures within these topoi are of physical significance. This is still very far from a theory of quantum gravity, but it can already shed some light on ordinary quantum theory, since we avoid the usual instrumentalism. Moreover, the formalism is general enough to allow for major generalisations. I will conclude with some more general remarks on related developments.