In the crowded courtyards of southern China's urban core, a research team has entrusted an evolutionary algorithm with a question that has long tested human judgment: how to make a small patch of green simultaneously beautiful, affordable, and climatically useful. By running thousands of simulated generations of plant selection and spatial arrangement, the NSGA-II algorithm found configurations that outpaced experienced designers on all three measures at once. The result is not merely a technical achievement but a quiet provocation — suggesting that in the geometry of constrained spaces, intel