Permanent deformation of metals results from the collective motion of a large number of dislocations. These dislocations are defects in the atomic lattice, and their motion causes the atomic layers to slide over each other. Many of these atomic-scale sliding events may combine to produce a macroscopic change of geometry that we observe as permanent deformation. Although good models are available both at the level of the atomic lattice and at the level of continuum mechanics, there currently is no method, formal or rigorous, that allows us to take a limit and connect the different scales in a convincing way. This is a major obstacle for the design and engineering of high-performance steels. In this talk I describe recent progress in this area, together with Lucia Scardia, Marc Geers, and Ron Peerlings. I will show how a simplified setup produces new insight into this difficult connection, and provides a hint as to how we should think about the more general case.