In this paper, we take one of the most emblematic models of the economic orthodoxy, the representative agent optimal growth problem, and discuss the adaptations it needs to go through to be reflective of a virtual world of interacting agents. The rational agent that maximizes intertemporal utility is replaced by a profusion of heterogeneous households, who are endowed with distinct productivity and confidence levels, who interact locally, and who make consumption-savings decisions based on a boundedly rational rule (a heuristic). We show that the three highlighted features (heterogeneity, local interaction, and non-optimality) are intertwined, and that the transformation of the standard optimal growth problem into a complexity framework requires their simultaneous consideration. Heterogeneous productivity levels trigger different technology absorption capabilities and, as a consequence, a slow process of innovation diffusion; the consumption heuristic introduces flexibility into consumption-savings choices, allowing for the coexistence of those who save with those who consume their entire current income; random contact across a population of agents makes sentiments of optimism or pessimism to spread in unpredictable ways. These processes tend to reinforce one another, provoking a change of scenery, with the conventional equilibrium growth model giving place to a multi-agent decentralized interaction platform where emergent results, rather than mechanic outcomes, are the norm. Ultimately, the new theoretical framework preserves the fundamental concept of what an economic growth model should be, at the same time it offers a richer structure of analysis, allowing for a deeper debate on the dynamics of the aggregate economy.