The epsilon-intelligent competitive equilibrium algorithm is a decentralized alternative to Walrus' tatonnement procedure for markets to arrive at competitive equilibrium. We build on the Gode-Spear-Sunder zero-intelligent algorithm in which random generation of bids and offers from agents' welfare-enhancing opportunity sets generates Pareto optimal allocations in a pure exchange economy. We permit agents to know if they are subsidizing others at such allocations, and to veto such allocations, restricting the subsequent iterations of the algorithm only to those trades that are both Pareto-improving and provide strictly greater wealth, and ultimately utility, for such agents. In this simple institution actions of minimally intelligent agents based on local information can lead the market to approximate competitive equilibrium in a larger set of economies than the tatonnement process would allow. This helps address one of the major shortcomings of the Arrow-Debreu-McKenzie model with respect to the instability of tatonnement in an open set of economies. It also addresses the behavioral critique of mathematically derived equilibria for the inability of cognitively-limited humans to maximize. The proof of convergence of the algorithm presented here also provides a way of showing the existence of competitive equilibrium for monotonic, convex exchange economies with heterogeneous agents and many goods without application of a fixed-point theorem.