Stable bacterial cross-feeding interactions, where one strain feeds on the waste of the other, are important to understand, as they can be a first step towards bacterial speciation. Their emergence is commonly observed in laboratory experiments using Escherichia coli as a model organism. Yet it is not clear how cross-feeding interactions can resist the invasion of a fitter mutant when the environment contains a single resource since there seems to be a single ecological niche. Here, we used digital organisms to tackle this question, allowing for detailed and fast investigations, and providing a way to disentangle generic evolutionary mechanisms from speci-ficities associated with E. coli. Digital organisms with evolv-able genomes and metabolic networks compete for resources in conditions mimicking laboratory evolution experiments. In chemostat simulations, although cross-feeding interactions regularly emerged, selective sweeps regularly purged the population of its diversity. By contrast, batch culture allowed for much more stable cross-feeding interactions, because it includes seasons and thus distinct temporal niches, thereby favoring the adaptive diversification of proto-species.