We study if adaptive learning by a Central Bank (CB) in the Kydland and Prescott environment can steer the economy to the Pareto-optimal outcome. Our CB evaluates its potential strategies regarding the announced and the actual inflation rate through expectations of the performance of these strategies, formed thanks to its mental model of the economy. This model is forward looking and adaptive at the same time. As a starting point, we follow Arifovic et al. (2010), and initially assume that there are two types of agents: Believers who set their inflation forecast equal to the announced inflation, and Non-believers who form static optimal forecast coupled with a forecast error correction mechanism. Our results show that the economy can reach near Ramsey outcomes most of the time. In the absence of Believers, the economies almost always converge to the Ramsey outcome.In their experiments with human subjects, Arifovic and Sargent (2003) showed that experimental economies reach and stay close to the Ramsey outcome most of the time, giving support to the ‘just do it’ policy recommendation. In light of the experimental findings, our model is of particular interest as it is the only agent-based or adaptive learning model that consistently selects the Ramsey outcome in this context.