Term Coding for Extremal Combinatorics: Dispersion and Complexity Dichotomies

We introduce \emph{Term Coding}, a novel framework for analysing extremal problems in discrete mathematics by encoding them as finite systems of \emph{term equations} (and, optionally, \emph{non-equality constraints}). In its basic form, all variables range over a single domain, and we seek an interpretation of the function symbols that \emph{maximises} the number of solutions to these constraints. This perspective unifies classical questions in extremal combinatorics, network/index coding, and finite model theory. We further develop \emph{multi-sorted Term Coding}, a more general approach in which variables may be of different sorts (e.g., points, lines, blocks, colours, labels), possibly supplemented by variable-inequality constraints to enforce distinctness. This extension captures sophisticated structures such as block designs, finite geometries, and mixed coding scenarios within a single logical formalism. Our main result shows how to determine (up to a constant) the maximum number of solutions $\max_{\mathcal{I}}(\Gamma,n)$ for any system of term equations (possibly including non-equality constraints) by relating it to \emph{graph guessing numbers} and \emph{entropy measures}. Finally, we focus on \emph{dispersion problems}, an expressive subclass of these constraints. We discover a striking complexity dichotomy: deciding whether, for a given integer $r$, the maximum code size that reaches $n^{r}$ is \emph{undecidable}, while deciding whether it exceeds $n^{r}$ is \emph{polynomial-time decidable}.