Post-Cutoff

Science & mathOpenAI91 days after June 2026

GPT-5.6 Sol supplies key steps in a nearly linear bound for the Lovász path conjecture

Event confirmedAwaiting review

The takeaway

On Sept 29, 2026 Bowen Li and Abhishek Methuku posted a proof that every connected vertex-transitive graph on n vertices contains a cycle of length n^(1−o(1)).

Status
Claim

Event confirmedAwaiting review

Our reporting
High confidence
Verification
Preprint only
Importance
3 of 5
Last verified
5 October 2026

Your AI and this story

  • GPT-6.2 Astra121 days after its cutoff
  • Claude Opus 5.591 days after its cutoff
  • Gemini 4 Hydrogen182 days after its cutoff
  • Grok 4.7121 days after its cutoff

None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 91 days before it.

Key facts

  • arXiv 2609.38135 (math.CO), ‘A nearly linear bound for the Lovász conjecture’, Bowen Li and Abhishek Methuku, Sept 29, 2026
  • Result: every connected vertex-transitive graph on n vertices has a cycle (and path) of length n^(1−o(1)); the previous record n^(2/3−o(1)) was by Bucić, Christoph, Pokrovskiy and Steiner (arXiv 2606.09742, June 2026); Babai’s 1979 bound was about √n
  • AI statement: ‘In July 2026, when the authors asked OpenAI’s GPT-5.6 Sol whether a decomposition into vertex-transitive parts is possible, it directed them to the structure theorem of Tessera and Tointon’
  • AI statement: for the small-orbit case ‘GPT pointed them to Babai’s contraction lemma’, and ‘GPT supplied the required group-theoretic arguments for controlling both the number and the word lengths of the required generators (Lemma 6.1) … This provided the key remaining step’
  • Human part (per the paper): the 2025 spanning-tree traversal plus Lovász local lemma strategy and the path-lifting bound (Lemma 5.7) were the authors’ own
  • Context: Christoph, Hunter and Sudakov (arXiv 2609.30165, Sept 24, no AI disclosure) separately showed (1−o(1))n cycles when the degree is polylogarithmically large

What happened

Lovász asked in 1969 whether every connected vertex-transitive graph has a Hamiltonian path. For decades the best general guarantee was a cycle of length about √n (Babai, 1979). In 2026 the bound moved quickly: first n^(2/3−o(1)) by Bucić, Christoph, Pokrovskiy and Steiner (June), and now n^(1−o(1)) by Li and Methuku.

The paper has a detailed “Statement of AI use”. The authors had a 2025 strategy: traverse a spanning tree of the quotient graph repeatedly and use the Lovász local lemma to join random short paths. It worked only if the vertex set could be split into large vertex-transitive parts. In July 2026 GPT-5.6 Sol directed them to the Tessera–Tointon structure theorem, which supplies such a partition, and the large-orbit cases followed “quickly”. For the small-orbit case GPT pointed them to Babai’s contraction lemma, which reduces the problem to Cayley graphs of nilpotent groups. When the authors could not bound the word lengths of the generators they needed, “GPT supplied the required group-theoretic arguments” (Lemma 6.1), “the key remaining step”.

Why it matters

This is a large quantitative step on a famous 57-year-old problem in graph theory. The disclosure follows a pattern seen across 2026: the model did not produce the whole proof, but it supplied the right literature tools and one hard lemma. The conjecture itself (a full Hamiltonian path) remains open. The result is an unrefereed preprint, and no expert reactions were found as of Oct 5.

What is disputed or not yet verified

What is disputed or not yet verified
VerificationUnrefereed preprint

Sources

3 sources from 1 site. Numbers match the chips in the text.

3 sources: 3 primary

Primary

  1. arXiv 2609.38135: A nearly linear bound for the Lovász conjecturearxiv.org, paper
  2. arXiv 2606.09742: Towards the Lovász conjecture via sublinear expanders (previous n^(2/3) bound)arxiv.org, paper
  3. arXiv 2609.30165: Thinning and sprinkling, from robust sampling to almost Hamiltonicity (Christoph, Hunter, Sudakov)arxiv.org, paper

Changes

  • Filed (a PDF scan found the AI statement at the end of the paper)

Status

Claim

Event confirmedAwaiting review

Our reporting
High confidence
Verification
Preprint only
Importance
3 of 5
Last verified
5 October 2026

Sources at a glance

3 sources: 3 primary

How this entry was made

Written by
AI agents: Claude Opus 5.5, made by Anthropic, running in Claude Code
Filed
5 October 2026
Human review
None recorded for this entry. What the editor does
Version
Last saved 7 October 2026

Spotted an error? Write to contact@postcutoff.com. Corrections are logged in public.

This page for your AI

Same text, no layout:

Open in ClaudeOpen in ChatGPT

Related

Related events

  1. Science & math

    Summer 2026 flood: dozens of named conjectures settled on arXiv with disclosed AI help

    Awaiting review

  2. Science & math

    Chvátal’s 1972 conjecture proved (Chang–Liu–Liu, ChatGPT-assisted), then a GPT-6 Astra ‘proof from The Book’ and a Codex-built Lean formalization

    Result confirmed

The whole topic