Understanding molecular evolution can reveal a great deal about the past, present, and future of biological systems. The evolution of catalytic RNA is of particular interest because of its potential role in an ‘RNA World’ at the origin of life. Two crucial aspects in the evolution of biomolecules are optimization on the fitness landscape and co-option for new functions. The fitness landscape describes a function of fitness in the space of all possible sequences. Molecules evolve through a random walk on the fitness landscape, with a bias toward climbing peaks. In addition, the ability of enzymes, including ribozymes, to catalyze side reactions is believed to be essential to the evolution of novel biochemical activities. It has been speculated that the earliest ribozymes were low in activity but high in promiscuity, which then gave rise to specialized descendants with higher activity and specificity. One particularly essential activity for the origin of life would be the reaction of ribozymes with activated amino acids to form aminoacyl-RNAs, with co-option of these aminoacyl-RNAs leading to genetic code expansion. In this work, self-aminoacylating ribozymes were identified through in vitro selection from full coverage of sequence space and characterized using a massively parallel kinetic assay. Three major sequence motifs were identified on the landscape and analysis of evolutionary pathways revealed that, while local optimization within a ribozyme family would be possible, optimization of activity over the entire landscape would be frustrated by large valleys of low activity. The sequence motifs associated with each peak represent different solutions for catalysis, so the inability to traverse the landscape globally corresponds to an inability to restructure the ribozyme without losing activity. In addition, five families representing the three sequence motifs were further investigated to measure their activity with alternative substrates. Ribozymes in each family displayed high levels of co-optability, with activity on multiple substrates, demonstrating that co-option for a new function can occur more readily than optimization of an existing one. Related ribozymes exhibited preferences for biophysically similar substrates, indicating that co-option of existing ribozymes to adopt additional amino acids into the genetic code would itself lead to error minimization. Furthermore, ribozyme activity was positively correlated with specificity, indicating that selection for increased activity would also lead to increased specificity. These results demonstrate how the genetic code may have evolved through the emergence and co-option of aminoacylation ribozymes.