Considerable experimental evidence has been collected on rules enhancing contributions in public goods dilemmas. These studies either confront subjects with prespecified rules or have subjects choose between different rule environments. In this paper, we completely endogenize the institution design process by asking subjects to design and repeatedly improve rule sets for a public goods problem in order to investigate which rules social planners facing a social dilemma ‘invent’ and how these rules develop over time. We make several noteworthy observations, in particular the strong and successful use of framing, the concealment of individual contribution information and the decreasing use of punishment.