The results reported upon in the paper demonstrate the feasibility of using an activation/inhibition dynamics among competence modules to solve the problem of action selection for an autonomous agent operating in a dynamic world. Such a scheme has particular advantages over traditional, deliberative hierarchical methods. The price to pay is that the actions selected might be less rational. However, the algorithm provides global controls which one can use to tune the action selection behavior along several criteria, such as thoughtfulness/rationality versus speed, goal-orientedness versus data-orientedness, and adaptivity versus bias to ongoing goals.