Remix.run Logo
▲ PhilipRoman 14 minutes ago

Assuming Lean kernel is correct (and there are multiple independent implementations of it, some proved in other ways themselves), the only part up for debate is whether the Lean-translated theorem (roughly half a page of code, maybe a bit more if you fully expand all the libraries it uses) accurately models the original GRH

The GRH is a bit too long to paste (but i encourage you to read it: https://github.com/google-deepmind/formal-conjectures/blob/m...), but for a sense of scale, here is the entire ABC conjecture:

  def radical (n : ℕ) : ℕ := n.primeFactors.prod id
  @[category research open, AMS 11]
  theorem abc (ε : ℝ) (hε : 0 < ε) :
      {(a, b, c) : ℕ × ℕ × ℕ | 0 < a ∧ 0 < b ∧ 0 < c ∧ ({a, b, c} : Set ℕ).Pairwise Nat.Coprime ∧
      a + b = c ∧ (radical <| a * b * c : ℝ)^(1 + ε) < c}.Finite := by
    sorry
not exactly beyond the real of human review