Scientific anti-realism is usually assumed to be a thesis about the scope of scientific theories with regard to unobservables. To many, this makes anti-realism an unattractive option, since it commits us to an arbitrary divide based on the limits of human perceptual organs and involves a skepticism about entities few want to reject. I argue that this view of what anti-realism should amount to comes from an inflationary and non-naturalistic meta-semantics that I argue should be rejected on independent grounds. I propose an alternative picture of anti-realism about science that draws no such arbitrary divide, but still helps us to dissolve the measurement problem--a problem that persistently resists realist solutions.