Overdiagnosis is difficult to classify, and thus to measure and to handle. From an etymological and technical point of view, overdiagnosis is an issue of diagnostics. It is obviously related to diagnosing a condition. However, the condition is correct (true positive). From the perspective of uncertainty (epistemology), overdiagnosis is not a question of uncertainty about the detected condition (accuracy), but about how this will develop (prognostic uncertainty). Therefore it relates more to treatment than to diagnostics. From an ethical point of view, overdiagnosis is special as it is a service offered to healthy people: it is like a treatment of health persons. Hence, overdiagnosis is like platypus. It does not fit our traditional categories, which makes it difficult to define, measure, and to handle. Therefore, overdiagnosis is a fascinating puzzle, not only to health policy makers, health professionals, and patients, but also to philosophers. Moreover, it raises the question of whether our conceptions of diagnosis and treatment are adequate, or whether we need new ways to look at medical activities. This presentation will use insights from the history of mathematics and the philosophy of science together with BigData and artificial intelligence (AI) to investigate whether overdiagnosis spurs a paradigm shift to ‘non-Euclidean’ medicine.