Abstract: New forms of machine-derived geographic information continue to draw the attention of geographers to unusual vistas on long-standing problems, and with those shifted vantages, new questions have presented for our models to address, at unusual scales and for novel geographies. In response, aspects of our discipline have “gone big”, seeking-out representative data and understanding over entire geographies and entire populations, with the goal of building comprehensive and holistic understanding of often complex systems. Meanwhile, a subset of inquiry is “going deep”, focusing on the minutia of geographic processes and phenomena in an effort to explore geographies in fine detail for fleeting moments of space and time. In this talk, I will discuss our work to apply geosimulation to “small geographies” in urban settings, with the goal of building new, explorative understanding of processes and phenomena that form from “atomic” units and relationships of (and within) human and built phenomena. In introducing this work, I will focus on three examples: modeling pedestrian mobility along streetscapes, simulating mass response to building collapse, and building gesture control for autonomous vehicles. In each case, challenges of understanding, representing, and modeling small geographies of urban settings come to the fore. Tackling challenges that manifest in these varied problem-sets has led us to the development of a new pipeline for geosimulation that we think holds significant promise for geographical inquiry at new scales of observation and understanding, and which offers new benefits for the development of geocomputation atop newly-forming data-sets, particularly those generated for and by spatially-aware machines.