I study a dynamic moral hazard problem, in which a principal hires an agent to develop and complete a project. The project is subject to failure risk and its progress is unobservable. The agent does not only control when the project is completed but also its propensity to fail or succeed upon completion. To motivate truthful disclosure of project failure, the optimal contract is lenient towards early but harsh towards late realizations of adverse events, such as project failure or delays in its completion. During the project’s development phase, the power of the agent’s incentives rises, not only leading to larger payments for later breakthroughs but also larger penalties for later failures. Overall, the agent is provided stronger equity incentives compared to the benchmark with observable project progress. For risky innovation projects, the optimal compensation scheme involves substantial tolerance for failure combined with a generous reward for success.