‘BioColony’ explores architecture as an assembly of biologically active materials, matters and spatial ecologies curating live-work-gardening typologies as a series of entanglements of species, genes and timescales. Building on the previous bio-receptive design approaches developed in RC7, Biocolony looks beyond solely the material condition and instead seeks to define new ways to plan building strategies to integrate living matter and ecologies into architecture. The project develops a design methodology using machine learning (ML) models trained on environmental data sets and a proposed platform tool that enables site-specific environmental information to be embedded into any given structure, thus optimising building mass and form for maximal growth of natural diversity. The typology becomes an inhabitable urban landscape of unruly territories that embraces ecological concepts of ageing, erosion and decay as a fundamental paradigm of resilience. Taxonomies of ecological parts are assembled using environmentally driven ML models to imagine how buildings as ‘biocolonies’ can serve as an ecosystem service for health.