Soar does not have to be completely programmed to work on a task. Although the task itself must be specified (as a problem space), Soar will use its default knowledge to begin work on the problem in the absence of any particular instructions about how to proceed. As this problem solving continues, chunks are built that cache the results of the problem solving. Thus, Soar's behavior in this situation is adaptive: it starts with little directly-applicable knowledge and gradually builds knowledge about task solutions. This is similar to Soar's movement from problematic to routine behavior. The emphasis here however is that Soar proceeds in the absence of any domain-specific knowledge; default knowledge imbues it with the `drive' to try to solve any problem given it.