A printed part is a solid shell around a sparse lattice, so most of its strength sits in the outer walls, and a bending stress is largest there too. Adding perimeters therefore does more than adding infill, and 100 % infill mostly adds weight and time [1]. The rules that follow reverse some habits of machined design: a thick section costs little when its inside is mostly empty, a square section beats an I-beam of the same envelope, and a lightening hole adds wall and print time. Material goes where the force flows, with fillets in the corners, as Stress concentration explains for any material. With partial infill the force over the outer area is an apparent stress, a property of the printed structure and not of the material, as the tensile test chapter states. Vase mode and beading patterns close the chapter: a single wall stiffened by its shape. Our data come from the cohort lab (one against six perimeters, four infill patterns at the same density) and from a bending test of shells against infill, the replication of CNC Kitchen’s comparison that the strategy lists first.