The ability of open crowds to produce useful and novel solutions is of significant research and practical interest. Much of the existing research has assumed that crowd participants approach problem-solving autonomously and independently of one another. In the Stack Exchange network, we explored problems posted by open crowds and the sequences of suggestions and comments leading up to a solution that was novel and judged by the solution seeker as useful for solving the problem. In the sequences of suggestions, we found indications of self-organization as the crowd emergently accumulated knowledge in specific patterns that were more likely to lead to accepted solutions. The patterns contained positive and negative forms of feedback ensuring that the crowd would, on its own and without external governance intervention, engage in actions that eventually led to accepted solutions. In total, based on the observed knowledge accumulation patterns, three self-organization principles were derived and incorporated into a theory of self-organization for creative solution generation in open crowds. Our theory development suggests implications for research on open crowds specifically, as well as for research on new organizational forms.