A random polynomial is a polynomial having random coefficients.
Let
where the
are independent and uniformly distributed on
. Kogler (2026) proved the
Odlyzko-Poonen conjecture that the probability that
is irreducible
over the rational numbers tends to 1 as
.
Kogler (2026) states that the proofs were generated by GPT-6 Astra, then rewritten and checked by the author. The paper also reports that Emmanuel Breuillard checked the proof and that GPT-6 Astra formalized the results and required background in Lean.
A random Littlewood polynomial instead has independently chosen coefficients in .
For a different model, let use the first
terms of a single infinite sequence of independent identically distributed real random variables. Assume
,
, and
for some
. If
counts the distinct real roots
of
, Chen et al. (2026) proved that the set
of subsequential limits of
is almost surely
where
is a universal constant defined by
. Here
,
with
,
is the upper-tail rate function for the zero count of the stationary Gaussian process
with covariance
. In particular
does not converge almost
surely. The lower limit
remains valid under the weaker condition
for some
. This sample-path assertion concerns coupled
successive polynomials, whereas the Kac
formula concerns an expectation.