The concepts of wide (or broad) and specific adaptation (or narrow) are intimately tied to genotype-by-environment interactions (GEIs) and have been widely, sometimes fiercely, debated among breeders and in the academia. Part of the debate has been fueled by the fact that wide adaptation was at the core of the Green Revolution. GEI is one of the most, if not the most often investigated issue in plant breeding because, when of qualitative type, it affects genetic gains by reducing heritability as shown by the breeder’s equation. The two components of GEI, namely Genotype by Locations (GL) and Genotype by Years within Locations (GYL), should be always estimated independently as they can be managed differently in a breeding program and it is the repeatability of GL that determines whether to breed for wide or for specific adaptation. The chapter argues that to address an heterogenous population of target environments, as it is the case of low input and organic agriculture, and to adequately address issues such as the client profile and the product profile, a breeding program based on decentralized selection, defined as selection in the target geographical and agronomic environment, can exploit GL interactions by subdividing the target population of environments into subgroups characterized by repeatable GL interactions and by selecting for specific spatial adaptation and wide temporal adaptation within each subpopulation. Such a breeding strategy is expected to reverse the trend of modern plant breeding toward a reduction of genetic diversity by increasing agrobiodiversity with beneficial effects both for the planet and for human health.