So how fast does the prediction model run? Will they soon be publishing predictions for millions of known sequences?