For me, I think the key was that much of my learning algebraic geometry was aimed at applying it somewhere else. Starting with a problem you know you are interested in and motivated about works very well. For me it was certain bits of geometric representation theory (which is how I ended up learning etale cohomology in the hopes understanding knots better), but for someone else it could be really wanting to understand Gromov-Witten theory, or geometric Langlands, or applications of cohomology in number theory.