We offer here a set of evidence-based optimal practices for social housing, applicable in general situations. Varying examples are discussed in a Latin American context. Adaptive solutions work towards long-term sustainability and help to attach residents to their built environment. We draw upon new insights in complexity science, and in particular the work of Christopher Alexander on how to successfully evolve urban form. By applying the conceptual tools of “Pattern Languages” and “Generative Codes”, these principles support previous solutions derived by others, which were never taken forward in a viable form. New methodologies presented here offer a promising alternative to the failures of the standard social housing typologies favored by governments around the world, which have proven to be dehumanizing and ultimately unsustainable.