Metallic glasses are characterized by two overarching physical principles: non-equilibrium and history-dependent behavior, and multiscale organization in time and space. This review examines recent advances in metallic glass research through this physical framework, spanning experimental characterization, additive manufacturing, atomistic and multiscale modeling, and machine learning and data-driven approaches. Recent experimental advances reveal how thermomechanical history is reflected in thermophysical properties, structure, dynamics, and mechanical response. Atomistic modeling provides microscopic insights into relaxation, crystallization, and deformation, while multiscale approaches connect local rearrangements to macroscopic behavior. Machine learning methods offer complementary tools for integrating data, identifying predictive descriptors, and supporting materials design. Additive manufacturing provides a particularly distinctive setting in which these two physical principles come together: repeated melting, rapid solidification, and spatially heterogeneous reheating generate structural states, defects, and property variations that depend sensitively on the local processing history. It therefore serves both as a promising practical route for fabricating complex metallic glass structures and as a stringent testbed for investigating how non-equilibrium processing and multiscale organization jointly control structure and properties. We conclude by highlighting open physics questions concerning non-equilibrium processing, deformation and failure, microscopic structure and dynamics, and the connection of mechanisms across different length and time scales.