A fundamental task of evolutionary biology is to explain the pervasive impression of organismal design in nature, including traits benefiting kin. Inclusive fitness is considered by many to be a crucial piece in this puzzle, despite ongoing discussion about its scope and limitations. Here we use individual-based simulations to study what quantity (if any) individual organisms become adapted to maximise when genetic architectures are more or less suitable for the presumed main driver of biological adaptation: namely, cumulative multi-locus evolution. As an expository device we focus on a hypothetical situation called Charlesworth’s paradox, in which altruism is seemingly predicted to evolve, yet altruists immediately perish along with their altruistic genes. Our results support a recently proposed re-definition of inclusive fitness, which is concerned with the adaptive design of whole organisms as shaped by multi-locus evolution, rather than with selection for any focal gene. They also illustrate how our conceptual understanding of adaptation at the phenotypic level should inform our choice of genetic assumptions in abstract simplified models.