A major constraint on experimental studies of geometrically frustrated magnets (GFMs) is the availability of suitable samples that provide tangible models to inspire or test theory. Most model GFMs studied to date have been ionic compounds, for which simple models of sphere packing predict an abundance of frustrated motifs—triangular plaquettes within close-packed layers, and nets of tetrahedra in three dimensions. The various frustrated topologies that arise naturally in such systems when they are decorated with moments and exchange interactions are reviewed. An alternative approach to the design of GFMs is to build networks of magnetic species from building blocks whose geometry or valence electron configuration lead to a strong preference for a particular connectivity. This strategy may be used to build relatively dense, mainly covalently bound networks of atoms or small molecules, or more open networks in which the building blocks are larger units. Most of the applications of this type of material have been in the field of catalysis so far, but they also present new opportunities to provide GFMs with rare or new lattice topologies.A phrase commonly quoted from the literature of cooking, meaning that before one prepares a dish, one must have the raw materials. It is also a lesson in reading the literature carefully: it does not actually occur in the text to which it is most commonly attributed .