I just read this press release from the university of Bath, claiming that the triplet genetic code has probably evolved from a doublet code.
Article abstract: "Explaining the apparent non-random codon distribution and the nature and number of amino acids in the standard genetic code remains a challenge, despite the various hypotheses so far proposed. In this paper we propose a simple new hypothesis for code evolution involving a progression from singlet to doublet to triplet codons with a reading mechanism that moves three bases each step. We suggest that triplet codons gradually evolved from two types of ambiguous doublet codons, those in which the first two bases of each three-base window were read (prefix codons) and those in which the last two bases of each window were read (suffix codons). This hypothesis explains multiple features of the genetic code such as the origin of the pattern of four-fold degenerate and two-fold degenerate triplet codons, the origin of its error minimising properties, and why there are only 20 amino acids." (Huan-Lin Wu, Stefan Bagby, Jean M.H. and den van Elsen (2005) J. Molec. Evol. Vol 61(1) p54-64)
I tought it is very interesting, but alas, I don't have access to the full text paper.