can a purely combinatorial object be approximated by a continuum geometry? i will describe evidence that the answer is "yes''  if that object is a transitive directed acyclic graph, otherwise known as a discrete order, otherwise known as a causal set. in which case, the approximating continuum geometry must be pseudo-riemannian with a "lorentzian'' signature of $(-, +, +, \ldots, +)$. i will, along the way, explain the crucial difference between riemannian and lorentzian geometry: in the former case the geometry is local and in the latter the geometry is, if not actually nonlocal then teetering on the edge of being nonlocal.  if there is time i will describe a model of random orders called transitive percolation, which is the lorentzian analogue of the erdős-renyi random graph and is an interesting toy model for a physical dynamics of discrete space-time.