Why Oatmeal Is Cheap: Kolmogorov Complexity and Procedural Generation

Procedural content generation in the gaming industry is fundamentally constrained by the relationship between a generator’s code size, its possibility space, and the Kolmogorov complexity of its output. While developers often scale the possibility space to create the illusion of vast content, this approach frequently results in the production of repetitive, unremarkable artifacts—a phenomenon colloquially known as the "Ten Thousand Bowls of Oatmeal" problem. The core thesis posits that meaningful variety is not a byproduct of scale, but rather a direct result of the design knowledge encoded into the generator. Because increasing the complexity of the underlying logic is significantly more labor-intensive than expanding the output space, systems often default to generating high volumes of low-information content.

This theoretical framework provides a rigorous mathematical basis for evaluating procedural systems across the global gaming sector. By establishing that the intricacy of generated assets is strictly bounded by the generator's design, the analysis highlights the inherent trade-offs between computational efficiency and qualitative depth. The study further formalizes these constraints by demonstrating how non-ideal, non-injective generators can be transformed into ideal systems with fixed input sizes. Through the mapping of inputs to unique binary strings, developers can ensure greater consistency and injectivity within their procedural pipelines.

Ultimately, the findings suggest that the industry must shift its focus from mere quantity to the density of design information. Achieving high-quality procedural generation requires a deliberate investment in the complexity of the generator itself rather than relying on simple algorithms that produce vast but redundant outputs. By applying these information-theoretic principles, developers can better navigate the limitations of their tools and create more meaningful, diverse, and engaging virtual environments.