Recent advances in sequencing technology have given rise to a new strategy in genome sequencing: whereas previously it was only possible to sequence and laboriously annotate one genome at a time, we now routinely sequence and compare groups of 5-10 phylogenetically related species and strains at a time. Does it still make sense to assemble whole genomes? We argue yes, on the condition that automatic whole-genome annotation and comparison can be performed on a large scale. We the results that can be obtained with our Magus comparative genomics pipeline, and show how its itegrated, comparative view of related genomes can be exploited for comparative whole-genome metabolic modeling and for high-level data-mining.