Over the last 20 years, probabilistic modeling has become the preeminent paradigm for complex statistical analysis in science and social science. Yet many probabilistic models remain difficult to represent, compose, estimate, and validate using traditional tools. This talk introduces Sense, a new data analysis environment that embeds probabilistic modeling into the core experience. The talk will demonstrate the power of this approach through a whirlwind tour of probabilistic model representation, composition, estimation, and validation. The talk will be targeted at both applied researchers in the sciences and social sciences and those interested in the future of statistical computing.