How do you actually build risky ideas inside a public company?