Separation of amino acids (AAs) in a mixture has been conventionally done with chromatography or electrophoresis; separation usually occurs over some continuous space. The present communication proposes a digital method for separating (and counting) single AA molecules in a mixture, with each of the 20 proteinogenic AAs ending up in its own discrete and spatially distinct container. Presently the method is designed for samples with a few molecules on up to the atto-mole level and can be used with samples collected from single cells. It is based on the superspecificity property of transfer RNAs (tRNAs): an AA can bind only with a cognate tRNA and not with any other; the binding error rate is about 1 in 350. Four necessary conditions for accurate separation are noted; it is shown informally that they are satisfied.