Abstract. This paper introduces AtomSwarm, a framework for soundbased performance using swarm dynamics. The classical ruleset for flocking simulations is augmented with genetically-encoded behaviours, hormonal flows, and viral `memes', creating a complex sonic ecosystem that is capable of temporal adaptation and self-regulation. The architecture and sound design methodologies are summarised here, with critical reference to its biomimetic design process, sonic spatialisation and self-organising capabilities. It is finally suggested that the system's lifelikeness is a product of its relational complexity, creating empathic enpurely through abstract formal structures.