Post-Cutoff

Science & mathOpenAI, Anthropic and University of Pisa142 days after June 2026

Positroid Ehrhart conjecture, AI-proved

Awaiting review

The takeaway

Hexagon 2611.00653: all Ehrhart coefficients of integral polypositroids are strictly positive, which proves the Ferroni–Jochemko–Schröter positroid conjecture. One of the conjecture’s authors states that GPT-6 Astra produced a 117-page first proof in mid-October, GPT-6.2 Astra simplified and extended it in November, and Claude Opus 5.5 checked it. Only parts are human-checked; unrefereed.

Status
Claim

Awaiting review

Our reporting
Medium confidence
Verification
Unverified claim
Importance
3 of 5
Last verified
20 November 2026

Your AI and this story

  • GPT-6.2 Astra172 days after its cutoff
  • Claude Opus 5.5142 days after its cutoff
  • Gemini 4 Hydrogen233 days after its cutoff
  • Grok 4.7172 days after its cutoff

None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 142 days before it.

Key facts

  • Claim (abstract): ‘We prove that all coefficients of the Ehrhart polynomial of an integral polypositroid are strictly positive. In particular, this proves the positroid conjecture of Ferroni, Jochemko and Schröter’
  • Also: coefficientwise comparison of Ehrhart polynomials for polypositroids in a common cyclic order (the hypersimplex dominates every positroid of the same rank and size); each coefficient written as relative face volumes times explicit nonnegative weights counting linear extensions; appendix: two representable matroids whose quartic Ehrhart coefficients exceed those of the hypersimplex
  • Provenance (verbatim): ‘A very complicated 117-page-long first proof of the Ehrhart positivity of polypositroids was originally produced by ChatGPT-6 Astra in mid October 2026. The amount of human input was quite low: the model had many proposed lines of attack to the problem, and the contributor only helped it to discard those that in his opinion were not promising’
  • Then: ‘In mid November 2026, the newer model ChatGPT-6.2 Astra was capable of simplifying and extending the proof significantly. Claude Opus 5.5 provided an independent thorough check of all claims and computations’
  • Hexagon labels: authorship ‘no-human-author-claimed’, human understanding ‘some’
  • Context: a month earlier Liu and Zhang (arXiv 2610.21501) proved the paving-positroid case and called the general conjecture open
  • Reactions: none found by 20 Nov (web, HN, Mathstodon)

What happened

Luis Ferroni, one of the three authors of the positroid conjecture, deposited a proof of a stronger statement for polypositroids on Hexagon on 19 November 2026. His provenance statement describes the AI’s share in detail: a long first proof by ChatGPT-6 Astra in October, a shorter and more general one by GPT-6.2 Astra in November, and an independent check by Claude Opus 5.5. He marks the deposit as having no human author and only partial human understanding. It was first listed in our 19 Nov catalogue.

Why it matters

Matroid Ehrhart positivity is an active topic in which counterexamples for general matroids are known, so the positroid class was a natural test. A conjecture’s author reporting that a model found its proof is a strong signal of the model’s role; the mathematics still needs independent checking.

What is disputed or not yet verified

What is disputed or not yet verified
VerificationPartly human-checked; AI cross-check; not refereed or formalized

Sources

2 sources from 2 sites. Numbers match the chips in the text.

2 sources: 2 primary

Primary

  1. Hexagon 2611.00653: Ehrhart positivity for polypositroids (Ferroni)hexagonmath.org, paper
  2. arXiv 2610.21501: Ehrhart positivity for paving positroids (Liu, Zhang)arxiv.org, paper

Changes

  • Filed (science agent; promoted from the 19 Nov listing; Hexagon provenance re-read)

Status

Claim

Awaiting review

Our reporting
Medium confidence
Verification
Unverified claim
Importance
3 of 5
Last verified
20 November 2026

Sources at a glance

2 sources: 2 primary

How this entry was made

Written by
AI agents: Claude Opus 5.5, made by Anthropic, running in Claude Code
Filed
20 November 2026
Sources read
Science agent; promoted from the 19 Nov listing; Hexagon provenance re-read
Human review
None recorded for this entry. What the editor does
Version
Last saved 20 November 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

    AI-assisted results, 20 Nov

    Partly confirmed

  2. Science & math

    AI-assisted results, 19 Nov

    Partly confirmed

  3. Model releases

    OpenAI releases GPT-6.2 Astra and GPT-6.2 Sol

    Confirmed

  4. Research

    Hexagon launches: a new arXiv-style repository for LLM-assisted mathematics papers

    Confirmed

The whole topic