News for hackers. Stuff that matters.
Re:Sharing AI progress in mathematics by adastra22 on Wednesday October 07, 2026 @12:16AM (#49988096)
> There's no Lean proof for this one

What is this then, vibes? Without a machine-checkable proof I'm not sure what to think of any of this.

Re:Sharing AI progress in mathematics by jboggan on Wednesday October 07, 2026 @12:29AM (#49988192)
Well I'm sure some people (maybe me if I had time) will do a write-up of this proof. It treads familiar ground for most of the setup, it's mostly the disk lemma and cancellation calculations that need to be understood, it's a fairly short paper and quite tractable.

I think it helps that basically everyone thinks this conjecture is true, it's just been so darn weird to attack. There's this odd thing that the induction proofs of this problem kept running into, which is that the N+1 condition would work except for in one tiny case when it could fail, but it would be covered by a very slightly stronger version of the conjecture. But then that would fail on one tiny case in induction, but you could solve that with another slightly stronger version. Etc., etc. I almost wondered if there were some sort of structure to the increasingly strong conditions and wanted to prove something about the meta-induction between the stronger conditions and the N's that they needed the next level to remain true. But that failed after 5 steps I think (Fable actually helped me write a few hundred test cases to explicitly show that pattern didn't continue forever, thank God).

BTW my existing test suite from previous proof attempts jives with this new algorithm, so I haven't seen any evidence yet that it's incorrect. Waiting for a Lean proof obviously.

Re:Sharing AI progress in mathematics by Daneel_ on Wednesday October 07, 2026 @12:44AM (#49988287)
It might have been updated. Is this the lean? https://github.com/openai/math/blob/main/lean/docs/180.md
Re:Sharing AI progress in mathematics by jboggan on Wednesday October 07, 2026 @12:58AM (#49988407)
Lol it should be, but it doesn't seem complete. Line 49 just says "sorry"

/-- Cubic bipartite three-vertex-connected plane graphs have a Hamiltonian cycle. -/ def MainStatement : Prop := ∀ (V : Type u) [Fintype V] [DecidableEq V] (G : SimpleGraph V) [DecidableRel G.Adj], G.IsRegularOfDegree 3 → G.IsBipartite → Planar G → ThreeVertexConnected G → HasHamiltonianCycle G

theorem main : MainStatement.{u} := by sorry

Re:Sharing AI progress in mathematics by SyzygyRhythm on Wednesday October 07, 2026 @01:33AM (#49988674)
In some cases they have a full Lean formalization; in others they just use it for the problem statement. Getting rid of that "sorry" means you've proved the statement. I'm not a Lean expert but it reads pretty clearly as the original conjecture (though the definition of PlaneEmbedding seems quite involved!).
Re:Sharing AI progress in mathematics by jboggan on Wednesday October 07, 2026 @01:57AM (#49988846)
I think this just has to be the problem statement, there's several lemmas I would expect to see in there. Granted I know very little about Lean but it seems like the question and not the proof outlined in the paper.
Re:Sharing AI progress in mathematics by jules on Wednesday October 07, 2026 @11:33AM (#49994288)
They are using a Lean tool where you separately state your theorems with `sorry` and then prove them elsewhere. The tool checks that all sorry's are covered. This is so the AI doesn't need to edit the specification of the theorem statement.
Re:Sharing AI progress in mathematics by jboggan on Wednesday October 07, 2026 @01:15PM (#49995794)
Ah, cool. I'm still learning Lean - is there somewhere else in the repo with the Lean specification of the cycle construction for the full argument?
Re:Sharing AI progress in mathematics by acomar on Wednesday October 07, 2026 @05:57PM (#49999337)
the json file next to the problem statement in lean says the solution starts here: https://github.com/openai/math/blob/main/lean/OAI/Combinator...

the proof is probably split over the constructions in the whole directory.

Re:Sharing AI progress in mathematics by jboggan on Thursday October 08, 2026 @12:58AM (#50001930)
Aha, thank you, this makes more sense now.