For many years, the holy grail of protein engineering has been the design of artificial amino acid sequences that fold into stable proteins with desired functions. In the current issue of Nature, two papers from the Ranganathan group (Russ et al., 2005; Socolich et al., 2005) report remarkable success in the design of artificial WW domains. Their method, termed statistical coupling analysis (Lockless and Ranganathan, 1999), does not use structural or physicochemical information but instead extracts information about essential patterns of amino acids from the evolutionary record.