A central question in the origin of life concerns whether primitive metabolites and catalysts interacted randomly in solution, as often envisaged, or whether they were arranged from the start in ordered metabolic complexes. The latter possibility would be consistent with the hypothesis of metabolite channeling in extant cells, which holds that intermediates in many pathways are transferred directly, without diffusion, between pathway enzymes. A model on this basis is proposed in which life originated autotrophically de novo in metabolic complexes organized on FeS2 (pyrite) mineral surfaces. Because metabolites and catalysts arose at specific sites in these complexes, they could interact specifically with neighbouring species in evolving pathways prior to the existence of protein enzymes with precise substrate binding sites. In successive stages, RNA catalysts and protein enzymes could be incorporated in these arrays. The overall process may be viewed as a molecular analogue of embryonic development, with the formation and positioning of each new component continuously transforming the whole. A corollary of the hypothesis relates to the evolution of translation and the genetic code. By virtue of channeling, biosynthetically related amino acids (e.g., aspartic acid and threonine) would have arisen close together in the complex. A second premise is that tRNAs with similar base sequences, and thus similar anticodons, were also clustered together in the complex and channeled to adjacent sites. As a result of these combined effects, tRNAs with similar anticodons would necessarily have been positioned close to, and thus more likely to have been charged with, metabolically related amino acids. This mechanism affords a new rationale for the observed codon structure of the genetic code, in which biosynthetically related amino acids possess similar codons.