The coding and catalytic capacity of RNA has raised the possibility that early, acellular life may have been RNA-based, but whether RNAs can generate complex catalytic networks is unclear. Vaidya et al. engineered different versions of a catalytic RNA (known as a ribozyme) from Azoarcus proteobacteria that all require covalent joining of the two constituent RNA fragments for full activity. Crucially, the authors adjusted the ribozyme specificities to catalyse either their own joining or the joining of a different ribozyme variant in a cyclical network. Experimental tracking of RNA catalysis and computational modelling showed that complex RNA networks can outcompete the 'selfish' autocatalysis and can spontaneously grow in complexity.