Agents with different computational limits need different semantic representations of the world; communication between them hits a hard threshold determined by capacity mismatch, and you can derive the minimum communication rate needed from the agents' capacity constraints alone.
This paper shows how agents with different computational capacities develop different 'semantic alphabets' when interacting with the same environment. It proves that communication between mismatched agents has a sharp threshold: below a critical rate, meaningful communication is impossible, but above it, information flows efficiently.