The climate crisis poses a problem of global collective action that governments have sought to address through high-level international negotiations. The resulting arrangements, such as the 2015 Paris agreement, mandate compulsory reductions in emissions and require comprehensive plans for meeting those targets. The authors argue that these treaties are failing because they provide inadequate opportunity for initiative and experimentation at the subnational level. Since the route to successful emissions reduction is fundamentally uncertain, experimentation with different technologies and approaches is essential. Given that the best course is unknowable at the outset, successful strategies are likely to be developed locally and contextualized to local needs. Instead of insisting that all countries must endorse the details of a global agreement, like-minded governments and firms should forge ahead with their own solutions. Sharing the results of their efforts—what works, what doesn’t—will then limit the costs and risks of adaptation to climate change, while trade taxes and reputational penalties can bring reluctant governments and firms on board. The authors point to the 1987 Montreal Protocol, which protects the ozone layer, as a successful example of this kind of experimental governance. Californians similarly might point to their state’s stringent fuel economy standards and to its plan for banning the internal combustion engine.