Crushing normal surfaces is a key tool for modern 3-manifold software: it lies at the heart of algorithms for problems like unknot recognition, 3-sphere recognition and connected-sum decomposition. Crushing works well for spheres and discs, but things become more difficult when we wish to cut up our 3-manifold along more complicated surfaces. For example, it is natural to work with annuli if we want to detect composite knots, Seifert fibre spaces with nonempty boundary, or product I-bundles over surfaces. In this talk, I will present a workaround for annuli. Roughly, the idea is to fill in boundary components of a 3-manifold with solid tori, so that the annuli of interest become spheres and discs that we know how to crush. We track the fillings using an object called an ideal loop. In joint work with Eric Sedgwick and Jonathan Spreer, this idea has led to a practical algorithm for deciding whether a knot is prime or composite, and we hope to extend the idea to design algorithms for other problems