A digital twin is defined as a virtual representation of a physical asset enabled through data and simulators for real-time prediction, optimization, monitoring, controlling, and improved decision-making. Unfortunately, the term remains vague and says little about its capability. Recently, the concept of capability level has been introduced to address this issue. Based on its capability, the concept states that a digital twin can be categorized on a scale from zero to five, referred to as standalone, descriptive, diagnostic, predictive, prescriptive, and autonomous, respectively. The current work introduces the concept in the context of the built environment. It demonstrates the concept by using a modern house as a use case. The house is equipped with an array of sensors that collect timeseries data regarding the internal state of the house. Together with physics-based and data-driven models, these data are used to develop digital twins at different capability levels demonstrated in virtual reality. The work, in addition to presenting a blueprint for developing digital twins, also provided future research directions to enhance the technology.