From its advent in 2005 within global climate change negotiations, reducing carbon emissions from deforestation and other forest‐related activities (so‐called REDD+) has been experimented with in developing country contexts for over a decade now, with a wide array of expectations coming to be associated with it. Three consecutive conceptualizations are identifiable: carbon‐centered, where REDD+ is primarily a climate mitigation strategy; co‐benefits‐centered, where REDD+ becomes a triple win solution for climate, biodiversity and communities; and landscape‐centered, where REDD+ activities are embedded in integrated sustainable land‐use approaches. In assessing such evolving expectations against existing REDD+ experiences, a mixed picture emerges. Some expectations, specifically relating to forest carbon financing, are not being adequately met, while others, notably the delivery of co‐benefits, hold out more promise. Yet this also highlights a potential paradox facing REDD+. While there is growing recognition that co‐benefit generation is key, and that piece‐meal, forest‐carbon focused REDD+ interventions are unlikely to address the complex causes of tropical forest loss, forest carbon is still being foregrounded in measuring and reporting on REDD+ performance, and in generating results‐based payments (even as these aspects remain challenging). This implies, however, that the future of REDD+ may lie not in one conceptualization coming to dominate, but rather in co‐existence of heterogeneous practices. REDD+ may end up as a patchwork of projects and practices with different foci and financing mechanisms. Although this cannot prevent trade‐offs, such a heterodox REDD+ may provide building blocks for the polycentric governance of the world's remaining tropical forests. WIREs Clim Change 2017, 8:e425. doi: 10.1002/wcc.425This article is categorized under: Climate and Development > Social Justice and the Politics of Development Policy and Governance > Multilevel and Transnational Climate Change Governance