A Steiner triple system is a collection of points and triples (three-element subsets) arranged so that every pair of points belongs to exactly one triple. These objects appear naturally in combinatorics and design theory, and mathematicians have long studied how to tell two such systems apart. The central question here is: how complicated is the problem of classifying countable Steiner triple systems up to isomorphism, meaning up to relabeling of points that perfectly matches one system to another?
The paper answers this by showing that classifying countable Steiner triple systems is as hard as it can possibly be within a certain precise framework, called Borel complexity theory. This framework provides a way to compare classification problems across all of mathematics by asking whether one classification problem can be systematically translated into another. Showing a class is "Borel complete" means its classification problem is at least as hard as classifying all countable mathematical structures whatsoever. The authors prove this by building an explicit, well-behaved procedure that takes any countable graph (a network of nodes and edges) and produces a Steiner triple system in a way that perfectly preserves the isomorphism type. Two graphs are structurally identical if and only if their corresponding Steiner triple systems are structurally identical.
What makes the result especially strong is that the construction also preserves symmetry: the automorphisms of the graph (the ways you can rearrange a graph while keeping its structure intact) correspond exactly to the automorphisms of the resulting Steiner triple system. This means no information is lost or distorted in the translation. The upshot is that there is no hope of finding a simple, complete set of numerical or combinatorial invariants that could classify all countable Steiner triple systems, because doing so would simultaneously solve an impossibly hard classification problem for all countable structures.