The codon capture model avoids the above difficulty by suggesting that in the lineages with low GC-content (such as Firmicutes) codon usage can be skewed to such a degree that the GC-rich codons become virtually extinct (“AT-pressure”). Subsequently, they can be “re-captured” by the non-cognate tRNAs through translational misreading errors, effectively evading selection pressure, and expand over time through genetic drift. The ambiguous intermediate model, in contrast, starts with the mutations in tRNA that lead to the ambiguous recognition of a codon by a cognate tRNA and a new mutant tRNA. The latter subsequently takes over through, again, genetic drift. Therefore, the ambiguous intermediate model is not “neutral,” as the ongoing codon reassignments are expected to have deleterious effects on the organismal level. However, the advantage of the ambiguous intermediate hypothesis is that it does not require initial disappearance of certain codons, although it does favor more rare codons for reassignment, due to the lesser gross changes of the proteome. Importantly, experimental evidence in support of both the codon capture hypothesis and the ambiguous intermediate hypothesis exists (in silico and in vivo studies of codon reassignments in Candida (yeast) species), suggesting that the two are not mutually exclusive. Among the alternative hypotheses proposed, “genome streamlining” is of a special interest. It adds the global optimization aspect to the codon reassignment phenomenon, postulating that the driving force behind the code evolution might be minimization of the translation apparatus.