Economic models of climate change often take the problem seriously, but paradoxically conclude that the optimal policy is to do almost nothing about it. We explore this paradox as seen in the widely used DICE model. Three aspects of that model, involving the discount rate, the assumed benefits of moderate warming, and the treatment of the latest climate science, are sufficient to explain the timidity of the model's optimal policy recommendation. With modifications to those three points, DICE shows that the optimal policy is a much higher and rapidly rising marginal carbon price; that higher carbon price has a greater effect on physical measures of climate impacts. Our modifications exhibit nonlinear interactions; at least at low discount rates, there is synergy between individual changes to the model. At low discount rates, the inherent uncertainty about future damages looms larger in the analysis, rendering long-run economic modeling less useful. Our analysis highlights the sensitivity of the model to three debatable assumptions; it does not, and could not, lead to a more reliably “optimal” cost of carbon. Cost-effectiveness analysis, focusing on the generally shorter-term cost side of the problem, reduces the economic paradoxes of the long run, and may make a greater contribution than economic optimization modeling.