Nucleic acid synthesis is precisely controlled in living organisms by highly evolved protein enzymes. The remarkable fidelity of information transfer realized between template and product strands is the result of both the spatial selectivity of the polymerase active site for Watson–Crick base pairs at the point of nucleotide coupling and subsequent proof-reading mechanisms. In the absence of naturally derived polymerases, in vitro template-directed synthesis by means of chemically activated mononucleotides has proven remarkably inefficient and error-prone. Nevertheless, the spontaneous emergence of RNA polymers and their protein-free replication is frequently taken as a prerequisite for the hypothetical ‘RNA world’. We present two specific difficulties that face the de novo synthesis of RNA-like polymers in a prebiotic (enzyme-free) environment: nucleoside base selection and intramolecular strand cyclization. These two problems are inherent to the assumption that RNA formed de novo from pre-existing, chemically-activated mononucleotides in solution. As a possible resolution to these problems, we present arguments and experimental support for our hypothesis that small molecules (referred to as ‘molecular midwives’) and alternative backbone linkages (under equilibrium control) facilitated the emergence of the first RNA-like polymers of life.