SUMMARYGlobal agriculture is dominated by a handful of species that currently supply a huge proportion of our food and feed. It additionally faces the massive challenge of providing food for 10 billion people by 2050, despite increasing environmental deterioration. One way to better plan production in the face of current and continuing climate change is to better understand how our domestication of these crops included their adaptation to environments that were highly distinct from those of their centre of origin. There are many prominent examples of this, including the development of temperate Zea mays (maize) and the alteration of day‐length requirements in Solanum tuberosum (potato). Despite the pre‐eminence of some 15 crops, more than 50 000 species are edible, with 7000 of these considered semi‐cultivated. Opportunities afforded by next‐generation sequencing technologies alongside other methods, including metabolomics and high‐throughput phenotyping, are starting to contribute to a better characterization of a handful of these species. Moreover, the first examples of de novo domestication have appeared, whereby key target genes are modified in a wild species in order to confer predictable traits of agronomic value. Here, we review the scale of the challenge, drawing extensively on the characterization of past agriculture to suggest informed strategies upon which the breeding of future climate‐resilient crops can be based.