There are indications that the current generation of models used to simulate the geography of housing choice has reached the limits of its usefulness under existing specifications. The relative stasis in residential choice modeling--and urban simulation in general--contrasts with simulation efforts in other disciplines, where techniques, theories, and ideas drawn from computation and complexity studies are revitalizing the ways in which we conceptualize, understand, and model real-world phenomena. Many of these concepts and methodologies are applicable to housing choice simulation. Indeed, in many cases, ideas from computation and complexity studies--often clustered under the collective term of geocomputation, as they apply to geography--are ideally suited to the simulation of residential location dynamics. However, there exist several obstructions to their successful use for these puropses, particularly as regards the capacity of these methodologies to handle top-down dynamics in urban systems. This paper presents a framework for developing a hybrid model for urban geographic simulation generally and discusses some of the imposing barriers against innovation in this field. The framework infuses approaches derived from geocomputation and complexity with standard techniques that have been tried and tested in operational land-use and transport simulation. As a proof-of-concept exercise, a micro-model of residential location has been developed with a view to hybridization. The model mixes cellular automata and multi-agent approaches and is formulated so as to interface with meso-models at a higher scale.