How do special number systems help mathematicians pack an unusually large number of point pairs exactly one unit apart?

This asks why CM fields sit at the center of recent constructions that pack an unusually large number of unit distances among points in the plane. The short version: CM fields supply the algebra that lets you manufacture many points at exactly distance 1 from each other. The newest work shows the trick gets much stronger when you let those number systems grow without limit. The corpus covers this with only two notes, so treat what follows as a doorway, not a survey.

Some background the notes take for granted. The classic way to get lots of unit distances is to treat points in the plane as complex numbers and look for many numbers of 'length' exactly 1. The Gaussian integers, the simplest CM field, are the standard starting point. A CM field generalizes this: it pairs a field of real numbers with a matching imaginary extension, so there is a well-behaved notion of complex conjugation, and with it a well-behaved notion of length. That structure lets you build a lattice of points and count how many pairs sit at distance 1. How many unit distances can points in a plane have? describes exactly this kind of construction. It combines CM fields with Golod–Shafarevich arguments, a classical tool for producing infinite towers of number fields, to get n points with more than n^1.014 unit-distance pairs. Erdős conjectured that the count could not grow meaningfully faster than n, so even an exponent slightly above 1 matters.

The part you might not expect: the breakthrough did not come from a new tool. What made OpenAI's unit distance counterexample succeed? reports that OpenAI's counterexample to the Erdős conjecture is built almost entirely from decades-old number theory. The new move was to let the degree of the field, roughly its 'dimension' over the rationals, grow to infinity instead of staying fixed. With growing degree, a single fixed prime can keep two quantities under control: the class number and the discriminant, which roughly measure how irregular and how 'large' the field is. Keeping those small is what lets the lattice stay dense enough to beat linear growth. So the CM field is the stage, the Golod–Shafarevich tower is the supply of ever-larger stages, and the growing degree is what makes the stages pay off.

The two notes also show a two-step pattern. OpenAI's result proved that superlinear growth exists but did not say by how much. The follow-up lattice construction made the exponent explicit at n^1.014, though the bound still carries an unspecified absolute constant. If you want the mechanism, start with the counterexample note. If you want the number, read the lattice note. The corpus does not go further into the algebraic number theory itself, such as why CM fields specifically rather than other fields, or how tight the exponent might get, so those questions need outside sources.