Software-intensive Science, and in particular the method of modelling large and complex systems by means of computer simulation, presents acute dilemmas of epistemic trust. Some authors have contended that simulations are essentially epistemically opaque vis and vis a human agent, others that complex simulation models suffer from an inescapable confirmation holism. We argue that the shortcomings lie in the failure of modellers to adopt sound Software Engineering practices, and that the elevation of computational models into superior epistemic authorities runs counter to principles that are common to both Science and Software Engineering.