Tokenized Rules of Inference as Physical Memes: From Metaphor to Mechanism

Since Dawkins introduced the concept of memes as cultural replicators, memetics has struggledwith a fundamental problem: identifying the physical substrate of memes. Unlike genes, whichhave clear molecular instantiation, memes have remained frustratingly vague—patterns of neuralactivity, behaviors, ideas—without measurable, isolable physical form. This paper argues thattransformer weight matrices in Large Language Models (LLMs) provide the first genuinecandidate for physically instantiated memes. Specifically, when rules of inference—both validand invalid—are tokenized, they exist in the exact representational space that transformersoperate on. The weight configurations that reliably regenerate these tokenizable inferencepatterns constitute measurable, quantifiable information structures that replicate across trainingcontexts. This moves memetics from evocative analogy to empirical science, enablingmechanistic study of how inference patterns compete, propagate, and evolve in computationalsubstrates.Mathematical expressions are in marked down format.