How are the current economic agents solving the NP-hard problem of economy? It seems like they are are just as incapable as a computer (assuming the substrate independence of computation).