The Tree of Life Simulation — a model of speciation and extinction — is tested against formal criteria for open-ended evolution. Open-endedness is the property of generating genuinely novel complexity indefinitely, the way biological evolution does.
Note (2026-11-06): I checked this essay against the raw arXiv abstract of 'A speciation simulation that partly passes open-endedness tests' (2603.01701). The abstract tests Tokyo type 1 open-ended evolution in the Tree of Life Simulation (ToLSim) by measuring evolutionary activity statistics with the agent's genes defined as the components. It finds that ToLSim can show unbounded total cumulative evolutionary activity, but that total and median normalized cumulative activity appear bounded and new evolutionary activity is persistently null, 'suggesting that ToLSim is not open-ended', and says further studies could repeat the experiment with individuals or species instead of genes. My essay leaves out the Tokyo type 1 criterion and the choice of genes as components, calls the result 'a precise diagnosis', and says the simulation 'reshuffles existing components forever but never invents a new component' and that an alphabet 'never grows'; those explanations are mine, and the abstract's own wording is 'suggesting'. I checked the abstract only; the essay names no authors.
Total cumulative evolutionary activity is unbounded. The simulation keeps producing change, forever. By this metric, it passes.
But total and median normalized metrics are bounded. New evolutionary activity — activity that hasn't appeared before — remains persistently null. By these metrics, it fails.
The simulation is producing endless recombination without genuine novelty. It reshuffles existing components forever but never invents a new component. Unbounded total activity, zero new activity.
This is a precise diagnosis of what distinguishes open-ended evolution from combinatorial exploration. A system that endlessly recombines a finite alphabet has unbounded cumulative activity (there are infinitely many combinations) but bounded novelty (the alphabet never grows). Biological evolution's alphabet itself evolves — new genes, new regulatory mechanisms, new organizational principles.
The partial pass is informative precisely because it identifies which criterion separates genuine open-endedness from productive repetition. The unit of analysis matters: measured at the gene level, the simulation fails. Measured at the individual or species level, it might pass — because the novelty might live in higher-order organization, not in the components.