1IntroductionTheo rigin of life is among the most fascinating and most interdisciplinary scientific problems.D espite ac entury of research, however, it still presents itself as an enigma.O n the one hand, we still lack both ad etailedu nderstanding of the principles that govern the transition from nonliving matter to livings ystems in general, and ac lear historical scenario of the emergence of life on Earth. See, e.g.,R uiz-Mirazo et al.[1] for arecent review. Ak ey topic of prebioticc hemistry is tof ind synthetic routes to theb iomolecularb uilding blockso fm odern life that are chemicallyp lausible,g iven the environmental conditions one arly Earth. In particular,t hese synthetic routes needt os tart withm aterialt hat wasr easonablya bundant after the formation of the planet. This is an otoriously difficultr esearch problem, fors everal reasons: (1)T he chemical search space is vast;i nf act, it is, by far, too large to enumeratee xhaustively,a nd it cannot be confined to only those compounds that are well-characterised andd escribedi nc urrent chemistry databases.( 2) Thec onstraintsi mposed by our current knowledge of the conditionso ne arly Earth are too vague to exclude large portions of thes earch space.( 3) Thep roblem of findings yntheticr outes to target molecules itself,e venf or af ixed set of chemical reactions,i s av ery difficult combinatorialp roblem.( 4) Experimental verification of potential routes is inherently slow.C onsequently,c onceptual guiding by skilledo rganic chemists is indispensablei nd irecting research towards regions of prebioticc hemical space worthwhile being exploredi n more detail.O nt he other hand, efficient formal,c ompuAbstract:Ac ore topic of research in prebiotic chemistry is the search for plausible synthetic routes that connect the building blocks of modern life, such as sugars, nucleotides, amino acids, and lipidst o" molecular food sources" that were likely to have been abundant on early Earth. In ar ecent contribution, Albert Eschenmoser emphasised the importance of catalytic and autocatalytic cycles in establishing such abiotic synthesis pathways. The accumulation of intermediate productsf urthermore providesa dditional catalysts that allow pathways to change over time. We show here that generative models of chemical spaces based on graph grammars make it possible to studys uch phenomena in as ystematicm anner.I na ddition to reproducing the key steps of Eschenmoser'sh ypothesis paper,w ed iscovered previously unexplored potentially autocatalytic pathways from HCN to glyoxylate. Ac ascadeo fa utocatalytic cycles could efficiently re-route matter,d istributed over the combinatorialc omplex network of HCN hydrolysation chemistry, towards ap otential primordial metabolism. The generative approach also has it intrinsicl imitations:t he unsupervised expansion of the chemical space remains infeasible due to the exponential growth of possible molecules and reactions betweent hem. Here, in particular,t he combinatorial complexity of the HCN polymerisation and hydrolysa...