Precise quantification of mental states is an area of intensive discovery and compelling need. Depression affects 405 million people globally, is our number one cause of disability and contributes to deaths by suicide occurring every 40 seconds. Yet, our current best efforts rely on a one-size-fits-all approach to detection and to treatment choices. We do not yet have a model that integrates experienced symptoms with underlying biology. This presentation outlines a new model for depression. Types of depression are identified by dysfunctions in large-scale human brain circuits. These types can improve the prediction of treatment outcomes as demonstrated in a large-scale biomarker trial. They offer a foundation for interpreting more distal sensors of behavior and clinical trajectories, and for extending to other areas of mental health and wellness. Experiences in developing this model, applying it in trials, and now expanding it with new computational approaches, sensors, interventions and real-world clinical applications are discussed.