The paper disproves a conjecture in a branch of mathematics that builds an analogy between ordinary number theory and the arithmetic of polynomials over finite fields. In classical number theory, a Wieferich prime is a special prime number satisfying an unusual divisibility condition. The paper works in an analogous setting where the integers are replaced by polynomials over a finite field, and ordinary multiplication is replaced by an operation called the Carlitz module. A "Carlitz-Wieferich prime" (c-Wieferich prime) is a polynomial prime satisfying the analogous special condition in this setting. In 2015, a mathematician named Thakur noticed from calculations and partial proofs that every c-Wieferich prime he could find had a degree (roughly, its "size") divisible by the characteristic p of the underlying field. He proposed this as a conjecture, and subsequent extensive computer searches seemed to support it.
The paper directly refutes this conjecture by exhibiting a concrete counterexample: an explicit c-Wieferich prime of degree 5 defined over a specific finite field with characteristic 19, where 19 does not divide 5. The authors also provide a clean closed-form mathematical description of the key quantity associated with this example, expressed through a degree-5 polynomial with coefficients in the simplest version of the field. Degree 5 turns out to be the smallest degree at which such a counterexample can possibly exist, which explains why earlier searches missed it.
The paper also reports that exhaustive computer searches find no counterexamples over the simplest "prime fields" across a wide range of degrees and characteristics, suggesting the conjecture may still hold in that restricted setting even though it fails in general. The method used to actually find the counterexample, along with proofs confirming the completeness of various searches, are described in a companion paper. The broader takeaway is that the analogy between classical prime numbers and polynomial primes over finite fields, while powerful and productive, has subtle limits that are difficult to predict from small amounts of data alone.