Attempts to implement autopoietic organization – essential for distinguishing the living from the non-living – in software and hardware models often miss key aspects, such as the realization of metabolic closure, which enable the system to interpret and interact with its physical environment. To address this, we aim to integrate chemical organization theory and active inference – both developed to encode the metabolism and cognition of autopoietic systems, respectively – to create a new generation of wetware models within the “Synthetic Biology (SB) and Embodied AI” research program. This approach may allow us to test critical aspects of synthetic life-like systems, including the implementation of sensory-motor loops and the anticipatory and predictive decision-making that arises from self-production and agent-environment interactions. Synthetic systems that exhibit operations beyond cybernetic (reactive) feedback loops and information processing could be considered minimal forms of chemically embodied cognition.