Procedural content generation (PCG) in games is often framed as a way to feed the content furnace, satisfying the voracious appetites of players by generating infinite seas of content for them to consume. Although this dominant framing provides a clear structuring purpose for PCG research, it also unnecessarily limits our vision of alternative purposes that generative methods might serve. Furthermore, generative systems designed with this purpose in mind may tend to reinforce certain problematic dynamics in game design. In this paper, we draw a contrast between two approaches to procedural terrain generation and the dynamics of play they tend to enable, which we term mining and gardening. We then extend this analysis to PCG more broadly and suggest that the latter (gardening) dynamic represents a viable and compelling alternative philosophy of how generative methods can be used in games.