Run multiple agents on the same task simultaneously, compare their approaches, and pick the best result. Like running a competition between Claude, Gemini, and Codex on the same problem.