The emergence of the genetic code remains an enigma. Proposed mechanisms are based on random, historical, thermodynamic and natural selection. However, they introduce chance as a key factor for overcoming the difficulties encountered by the model. We propose here a model in which three successive levels of chemical specificity generated the nucleotide assignments of amino acids in the genetic code. The first level results from hydrophobic and stereospecific interactions between amino acids and short oligonucleotides (termed oligons). The second and third levels of specificity are determined by conditions of energy transfer from loaded oligons (amino acid-oligomer covalently linked) to formation of phosphodiester bond (second level of specificity) and peptidic bond (third level of specificity), while these reactions are catalyzed by RNA templates. This model is sustained by the relationships observed between dipole moments of the nucleotides (forming the anticodon) and reactivity of the amino acyl linkage of the loaded oligon. Moreover, analysis of modern tRNAs reveals that they were probably generated by loose duplication of the nucleotide sequence forming the oligons, after emergence of the 'genetic code.' Indeed, the similarity of nucleotide composition with that of the anticodon decreases with the tRNA domain's distance from the anticodon, but the acceptor stem is relatively more similar to the anticodon than other stems closer to it. This would be because energy transfer constraints that existed between anticodon and amino acid in prebiotic loaded oligonucleotides still affect the structures of modern tRNA acceptor stems. In the model presented, the genetic code is inherent to the most archaic 'molecular physiology' in protolife, even before emergence of a functional 'protein world.' Simple physical processes, in which a level of specificity is integrated in an emerging meta-structure expressing new properties, generate a parsimonious and realistic explanation of emergence of the genetic code.