Boron's unique chemical properties and its reactions with metals have yielded the large class of metal borides with compositions ranging from the most boron-rich YB (used as monochromator for synchrotron radiation) up to the most metal-rich NdFeB (the best permanent magnet to date). The excellent magnetic properties of the latter compound originate from its unique crystal structure to which the presence of boron is essential. In general, knowing the crystal structure of any given extended solid is the prerequisite to understanding its physical properties and eventually predicting new synthetic targets with desirable properties. The ability of boron to form strong chemical bonds with itself and with metallic elements has enabled us to construct new structures with exciting properties. In recent years, we have discovered new boride structures containing some unprecedented boron fragments (trigonal planar B units, planar B rings) and low-dimensional substructures of magnetically active elements (ladders, scaffolds, chains of triangles). The new boride structures have led to new superconducting materials (e.g., NbRuB) and to new itinerant magnetic materials (e.g., NbFeIrB). The study of boride compounds containing chains (Fe-chains in antiferromagnetic ScFeRuB), ladders (Fe-ladders in ferromagnetic TiFeRhB), and chains of triangles (Cr chains in ferrimagnetic and frustrated TiCrIrB) of magnetically active elements allowed us to gain a deep understanding of the factors (using density functional theory calculations) that can affect magnetic ordering of such low-dimensional magnetic units. We discovered that the magnetic properties of phases containing these magnetic subunits can be drastically tuned by chemical substitution within the metallic nonmagnetic network. For example, the small hysteresis (measure of magnetic energy storage) of TiFeRhB can be successively increased up to 24-times by gradually substituting Ru for Rh, a result that was even surpassed (up to 54-times the initial value) for Ru/Ir substitutions. Also, the type of long-range magnetic interactions could be drastically tuned by appropriate substitutions in the metallic nonmagnetic network as demonstrated using both experimental and theoretical methods. It turned out that Ru-rich and valence electron poor metal borides adopting the TiCoB or the ThFe structure types have dominating antiferromagnetic interactions, while in Rh-rich (or Ir-rich) and valence electron rich phases ferromagnetic interactions prevail, as found, for example, in the ScFeRuRhB and FeRhRuB series. Fascinatingly, boron clusters (e.g., B rings) even directly interact in some cases with the magnetic subunits, an interaction which was found to favor the Fe-Fe magnetic exchange interactions in the ferromagnetic NbFeIrB. Using less expensive transition metals, we have recently predicted new itinerant magnets, the experimental proof of which is still pending. Furthermore, new structures have been discovered, all of which are being studied experimentally and computationally with the aim of...