The paper tackles a question in algebraic geometry about finding "nice" points on curves. Specifically, if you have a curve drawn in two-dimensional space over a very large algebraically closed field, and that curve is genuinely complicated in the sense that it cannot be described using only algebraic numbers, then you can always find a point on it whose two coordinates are as independent from each other as possible, meaning neither coordinate satisfies any polynomial equation involving the other. The paper also shows that if you have two such curves that are related to each other by a many-to-many algebraic correspondence, you can find a pair of points, one on each curve, that are simultaneously "nice" in this sense. These results are about locating points that carry as much independent information as possible.
The second part of the paper applies this geometric result to build exotic number systems called real closed fields. Real closed fields are algebraic structures that share the core properties of the real numbers, such as every positive element having a square root and every odd-degree polynomial having a root, but which can differ from the familiar real numbers in important ways. The paper constructs non-Archimedean real closed fields, meaning versions where there are numbers larger than any ordinary integer, with a precisely controlled degree of complexity measured by something called transcendence degree.
The most striking feature of the constructed fields is their rigidity: they have no non-trivial automorphisms, meaning the only way to map such a field to itself while preserving all arithmetic structure is to do nothing at all. This is a strong symmetry-breaking property. The paper achieves this for fields of every transcendence degree between 2 and the first uncountable infinity, filling in a range of cases that were not previously known to exist. The geometric result about finding independent points on curves is the key tool that lets the authors carefully control which elements of these fields can be mapped to which, ruling out any unexpected automorphisms.