Computational models of ecosystem development are essential tools for ensuring sustainable fishing, forestry, agriculture, and urbanization. In this project we develop a computational model for ecosystems populated by animals, plants, mountains, lakes, etc. The animal model is generic in the sense that its fundamental mechanisms of learning and decision-making are the same for insects, fish, birds, and mammals. The animals live in an environment that generates a never-ending stream of sensory data and respond by selecting actions at every point in time. Each animal has a notion of homeostasis and the goal of its learning and decision-making processes is to maintain homeostasis and thus survive as long as possible. At the core of the model is a reinforcement learning or genetic algorithm -one per animal or population. In experiments we show how the animal models learn to balance several needs, in particular energy, water, temperature, and social proximity. We also use the model to reproduce typical predator-prey dynamics.