For now, the best way to select a child's genes is to select a potential child who has those genes, using genetic testing and either selective abortion, sperm and egg donors, or selecting embryos for implantation. Some people even wish to select against genes that are only mildly undesirable, or to select for superior genes. I call this selection drift– the standard for acceptable children is creeping upwards. The President's Council on Bioethics and others have raised the parental love objection: (...) Just as we should love existing children unconditionally, so we should unconditionally accept whatever child we get in the natural course of things. If we set conditions on which child we get, we are setting conditions on our love for whatever child we get. Although this objection was prompted by selection drift, it also seems to cover selecting against genes for severe impairments. I argue that selection drift is not inconsistent with the ideal of unconditional parental love and, moreover, that the latter actually implies that we should practise selection drift – in other words, we should try to select potential children with the best genetic endowments. My endowment argument for the second claim works from an analogy between arranging an endowment prior to conception to fund a future child's education, and arranging a genetic endowment by selecting a potential child who already has it, where in both cases the child would not have existed without the endowment. I conclude with some programmatic remarks about the nonidentity problem. (shrink)