Disordered materials surround us in our daily lives from glasses and plastics to sandcastles and mudslides; however, their properties are poorly understood compared to their crystalline counterparts. Here we address two longstanding questions about glassy materials: how the microscopic structural defects in these systems lead to macroscopic mechanical properties and the causes of anomalous changes in the glass transition upon strong confinement. Until recently, defects in amorphous solids were elusive; however, a novel machine learning method for uncovering them has been discovered. This method allows for the construction of a structural quantity, softness, related to the rearrangement probability of a particle. We begin by comparing rearrangement size of to defect size, the softness correlation length, in a broad set of experimental and simulated materials. These length scales are similar suggesting that softness sets the rearrangement length scale in these materials. Next, we find these defects react similarly for a given amount of strain across all studied materials. Thus, a build-up of softness may set the universal yield strain in disordered materials. To better understand material failure, we introduce an ensemble of simulated polymer nanopillars that we strain apart. We build a machine learning model to detect where shear bands will form using structural features measured prior to deformation. We find that small density fluctuations at the pillar’s surface are the most important structural features to determine where shear bands form up to approximately 100nm in diameter. The importance of a plane’s mean softness to shear band classification grows with diameter suggesting softness predicts shear banding in the bulk. Finally, we turn to examining the dynamic heterogeneity of glassy films under strong confinement. To understand this, we bias model polymer thin film trajectories toward the dynamic first-order phase transition between high and low mobility dynamic basins. Changes in the transition under confinement are reminiscent of capillary condensation. The changes parallel changes in the glass transition observed in ultrathin film experiments suggesting a possible link between the effects we see and experiments.