We study the optimal allocation of tasks over the life cycle of an innovative project. Innovation requires two complementary activities, research and development, both subject to moral hazard. Agents also privately observe the quality of discoveries, so the principal must elicit truthful reports to avoid costly mistakes. Assigning both tasks to a single "solo operator" avoids free-riding but creates a double-deviation problem: the agent can shirk on research and conceal the resulting low-quality discovery. As time passes without a discovery, this deviation becomes increasingly attractive. Separating the tasks across two agents eliminates the double deviation but introduces free-riding. We characterize how this tradeoff shapes organization over time. When low-quality discoveries are readily exposed, the project is run by sequential solo operators; when they are difficult to expose, a two-agent team is optimal from the outset. At intermediate levels of detectability, the project begins with a solo operator and switches endogenously to a team.