How to use an incomplete and uncertain model of the environment to perceive, infer, decide and act efficiently? This is the challenge both living and artificial cognitive systems have to face. Logic is by nature unable to deal with this question. The subjectivist approach to probability is an alternative to logic specifically designed to face this challenge. In this paper we introduce Bayesian Programming, a methodology, a for- malism and an inference engine to build and compute probabilistic models. The principles are illustrated with two examples: modeling human perception of structure from motion and playing to train a video game avatar.