This is the unforced version of the problem. OpenAI found an example that uses a carefully designed external force to break the equations, so this one is more "realistic" in a sense.