Spatial AI / world models

Spatial AI refers to machine-learning systems that build internal 'world models' — learned, predictive representations of a physical environment's structure and dynamics — rather than reasoning from text alone. In ATH it is the forward-looking layer that lets an AI reason over the building as a three-dimensional, time-evolving habitat (the Anthropogenic Cliff-Face), predicting where colonisation and degradation will propagate. It connects the geospatial pathology mapping and digital-twin data to autonomous, spatially-aware stewardship.