---
id: "2026-10-06-openai-deligne-drinfeld-conjecture-proof"
url: "https://postcutoff.com/e/2026-10-06-openai-deligne-drinfeld-conjecture-proof/"
as_of: "2026-10-30T13:42:00+01:00"
date: "2026-10-06"
date_precision: day
category: science
importance: 5
confidence: high
status: [Result confirmed]
verification: Lean-verified
sources: 13
editor: human editor
human_review: null
version: "2026-10-29"
---

As of: 2026-10-30 13:42 CET.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/
Canonical page: https://postcutoff.com/e/2026-10-06-openai-deligne-drinfeld-conjecture-proof/

# Deligne–Drinfeld conjecture proved; experts simplify and reprove OpenAI's proof

Full title: Deligne–Drinfeld conjecture: OpenAI's AI-generated proof is accepted and simplified by Thomas Willwacher, and Minoru Hirose gives an independent second proof

"The Deligne–Drinfeld conjecture" (OpenAI, 23 Sept 2026, released 6 Oct, family 008) proves that the rational Grothendieck–Teichmüller Lie algebra grt₁ with the Ihara bracket is freely generated by one element in each odd weight ≥ 3, also after completion. Brown (2012) and Willwacher (July 2026) had shown that such generators are free. The missing part was that they span all solutions of the defining equations, which computations had confirmed only through weight 29. OpenAI's 41-page proof bounds the dimensions through a degeneration of the pentagon equation in characteristic two. Willwacher (arXiv 2610.15592, 11 Oct, v3 22 Oct) confirms that this settles the conjecture and gives a simplified proof over 𝔽₂. Hirose (arXiv 2610.23719, 16 Oct) gives an independent route via a ℤ₍₂₎-integral Hoffman basis of multiple zeta values. The Lean main statement was accepted in the community Comparator run of 25–26 Oct.

## Key facts
- Claim (abstract, verbatim): 'We prove the Deligne–Drinfeld conjecture: the rational Grothendieck–Teichmüller Lie algebra is freely generated by one element in every odd weight at least three, with the corresponding isomorphism after weight completion.'
- What was open (paper, §1.2): Ihara and Drinfeld gave the odd-weight elements; Brown's theorem on mixed Tate motives over ℤ (2012) supplies a free Lie subalgebra; Willwacher (arXiv 2607.18793, July 2026) proved freeness for any family with a nonzero x²ᵏy coefficient; 'Freeness leaves open whether there are additional solutions of the defining equations', and Naef–Willwacher computations established equality 'through weight 29'. OpenAI's 'main estimate is the all-weight upper bound' dim W_n ≤ dim Lie_ℚ⟨e₃, e₅, e₇, …⟩_n
- Method (paper, §1.3): 'The main step is a degeneration of the pentagon in characteristic two', in letters A = x², C = [x, y], B = y, followed by integral deletion operators that specialize Hirose–Sato's algebraic differential operators. A stated consequence: H⁰(GC₂) of Kontsevich's graph complex is the completed free Lie algebra on generators of weights 3, 5, 7, … (via Willwacher's 2015 isomorphism)
- Thomas Willwacher, 'The Deligne–Drinfeld conjecture after OpenAI' (arXiv 2610.15592; v1 11 Oct, v3 22 Oct): 'The remaining part of the Deligne–Drinfeld conjecture was solved by OpenAI [Ope26]'. He gives 'a streamlined and significantly simplified version of the proof of OpenAI. The author claims no originality regarding the overall proof strategy', and proves a stronger statement in characteristic 2 (grt₁(𝔽₂) is freely generated by explicit elements σ₂ₘ₊₁). He credits Leila Schneps and Benjamin Enriquez, whose 'reworked version of the OpenAI proof inspired the simplified formula for the representation ρ'. AI statement: AI tools were 'used extensively in all stages of this work, helping the author in understanding, restructuring and simplifying the original arguments'
- Minoru Hirose (Kagoshima), 'Confluence relations, Hoffman basis over ℤ₍₂₎, and the Deligne–Drinfeld conjecture' (arXiv 2610.23719, 16 Oct): every multiple zeta value is a ℤ₍₂₎-combination of Hoffman-basis elements, as Hirose and Sato conjectured. Combined with Brown's theorem and Furusho's confluence characterization, this 'identifies the de Rham motivic Galois group of mixed Tate motives over ℤ with GRT and gives an alternative proof of the Deligne–Drinfeld conjecture'. 'Both the OpenAI proof and the present argument use reduction modulo 2.' Use of AI: ChatGPT (GPT-6 Astra Pro) 'for mathematical discussions and assistance with the exposition'; 'All mathematical arguments were independently verified by the author'
- Trinity College Dublin (15 Oct): a Hamilton Mathematics Institute workshop on Grothendieck–Teichmüller theory, planned long before the release, examined the proof. Leila Schneps (Paris) 'presented her understanding of the newly released AI-generated proof' on the first day, and Willwacher 'identified two new ideas that reduce the original, approximate 40-page proof to an argument of approximately eight pages'. Anton Alekseev (Geneva): 'A problem that had resisted mathematicians for many years has become the subject of an unexpected AI-generated argument.' Florian Naef (Trinity): mathematicians 'remain essential to interpreting, validating and developing the ideas that emerge'
- Lean (scope note 008): the formalization covers 'the rational solution space of antisymmetry, the three-term equation, and the four-strand pentagon with the Ihara bracket' and 'the corresponding continuous isomorphism after completion by weight'; 'The regularized-transport proposition and graph-complex consequences are not included.' The DeligneDrinfeld challenge was accepted in oai-math-revised's Comparator run of 25–26 Oct
- Other reactions: Christian Haesemeyer on X (20 Oct): the proof, now 'reduced to 3 pages', 'works by writing down a representation that comes out of nowhere, and then you just check well defined'. Hirose on X (21 Oct, in Japanese) says his result, combined with Brown's and Furusho's work and results on motives, also gives an alternative proof of the conjecture announced by OpenAI. Willwacher spoke 'On the Deligne–Drinfeld conjecture' in the online AHGT seminar on 22 Oct

## What happened

The Grothendieck–Teichmüller Lie algebra grt₁ is the infinitesimal symmetry algebra of the braiding and reassociation rules behind
Drinfeld's associators. It sits where knot theory, quantum groups, multiple zeta values and the motivic Galois group of mixed Tate
motives meet. Deligne and Drinfeld expected it to be as simple as possible: a free Lie algebra with one generator in each odd weight.
One half was known. Brown's 2012 theorem on mixed Tate motives gives such free generators, and in July 2026 Willwacher showed that any
reasonable choice of them is free. What remained was that these generators span every solution. Computer checks had confirmed this
only up to weight 29.

OpenAI's 6 Oct release contains a 41-page proof of that missing upper bound, by reducing the defining pentagon equation modulo 2. A
workshop on exactly this subject, planned long before, opened at Trinity College Dublin's Hamilton Mathematics Institute the following
week. Leila Schneps presented her reading of the AI proof on the first day. Thomas Willwacher, whose July paper had extended
Brown's half, posted a short version on 11 Oct and wrote that the remaining part of the conjecture "was solved by OpenAI". He
revised it twice and credits a reworked version by Schneps and Benjamin Enriquez. On 16 Oct Minoru Hirose posted a different proof
through the arithmetic of multiple zeta values. Both arguments, like OpenAI's, work modulo 2.

## Why it matters

Like Brennan's conjecture, this is a famous result from OpenAI's release that the people best placed to judge it have accepted. A
leading specialist states in print that the AI solved the remaining part, the proof has been cut to a few pages, a second independent
proof exists, and the main statement is machine-checked. The reactions have moved on to why the argument works: Haesemeyer says the
short proof rests on a representation that "comes out of nowhere", and Naef says the value of research also lies "in understanding
why it is true". None of the papers had been refereed as of 29 Oct.

## Your AI and this story

- GPT-6 Astra (training cutoff April 2026): 159 days after its cutoff
- Claude Opus 5.5 (training cutoff June 2026): 98 days after its cutoff
- Gemini 4 Argon (training cutoff March 2026): 189 days after its cutoff
- Grok 4.7 (training cutoff May 2026): 128 days after its cutoff

## Sources

1. [OpenAI: The Deligne–Drinfeld conjecture (23 Sept 2026, PDF)](https://github.com/openai/math/blob/main/preprints/The-Deligne-Drinfeld-conjecture-September-23-2026/paper.pdf) (github.com, paper)
2. [openai/math: Lean scope note for family 008](https://github.com/openai/math/blob/main/lean/docs/008.md) (github.com, code)
3. [openai/math: comparator challenge DeligneDrinfeld.lean](https://github.com/openai/math/blob/main/lean/ComparatorChallenges/DeligneDrinfeld.lean) (github.com, code)
4. [arXiv 2610.15592: The Deligne–Drinfeld conjecture after OpenAI (Willwacher)](https://arxiv.org/abs/2610.15592) (arxiv.org, paper)
5. [arXiv 2610.23719: Confluence relations, Hoffman basis over ℤ₍₂₎, and the Deligne–Drinfeld conjecture (Hirose)](https://arxiv.org/abs/2610.23719) (arxiv.org, paper)
6. [arXiv 2607.18793: On free generators in the Grothendieck–Teichmüller Lie algebra (Willwacher, July 2026)](https://arxiv.org/abs/2607.18793) (arxiv.org, paper)
7. [oai-math-revised: Comparator run of 25–26 Oct (results.tsv)](https://github.com/mathcommunity/oai-math-revised/blob/main/checks/comparator-2026-10-26/results.tsv) (github.com, code)
8. [Trinity College Dublin: Leading mathematicians gather in Trinity to examine new AI-generated mathematical proof (15 Oct 2026)](https://www.tcd.ie/news_events/articles/2026/delignedrinfeld-conjecture-workshop/) (tcd.ie, press)
9. [HMI workshop 2026 schedule (Leila Schneps: The OpenAI proof of the Deligne-Drinfeld conjecture)](https://sites.google.com/view/hmi-workshop-2026/schedule) (sites.google.com, discussion)
10. [oai-math-revised: community notes on the Deligne–Drinfeld manuscript](https://github.com/mathcommunity/oai-math-revised/blob/main/preprints/The-Deligne-Drinfeld-conjecture-September-23-2026/AGENT_NOTES.md) (github.com, discussion)
11. [Minoru Hirose on X: alternative proof of the Deligne–Drinfeld conjecture (21 Oct 2026, Japanese)](https://x.com/hirose_minoru/status/2112725404645826716) (x.com, discussion)
12. [Christian Haesemeyer on X: 'a representation that comes out of nowhere' (20 Oct 2026)](https://x.com/CHaesemeyer/status/2112478250261946595) (x.com, discussion)
13. [AHGT seminar: Thomas Willwacher, On the Deligne–Drinfeld conjecture (22 Oct 2026)](https://ahgt.math.cnrs.fr/activities/workshops/Seminar-Taming_the_flow_2026/) (ahgt.math.cnrs.fr, discussion)

## Changes

- 2026-10-29 (filed): Created (manuscript PDF, DeligneDrinfeld.lean, formalization.yaml, oai-math-revised AGENT_NOTES and Comparator results, Willwacher's and Hirose's arXiv abstracts and PDFs, TCD news, HMI schedule, AHGT seminar page, Haesemeyer's and Hirose's X posts via fxtwitter; web and HN searches found no further expert comment)

## Related

- 2026-10-24: [oai-math-revised: mathematicians launch a community hub to revise OpenAI's AI-written math manuscripts (guest post on Tao's blog)](https://postcutoff.com/e/2026-10-24-oai-math-revised-community-hub/index.md)
- 2026-10-19: [AI-assisted results in the 19 Oct 2026 arXiv listing](https://postcutoff.com/e/2026-10-19-ai-assisted-results-october-19-listing/index.md)
- 2026-10-16: [Brennan's conjecture: experts confirm and shorten OpenAI's proof](https://postcutoff.com/e/2026-10-16-brennan-conjecture-openai-proof-binder-ivrii-prause-smirnov/index.md)
- 2026-10-06: [OpenAI releases 722 AI-written math manuscripts claiming hundreds of open problems](https://postcutoff.com/e/2026-10-06-openai-math-release-722-manuscripts/index.md)
- Archived: 2 posts, listed in https://postcutoff.com/e/2026-10-06-openai-deligne-drinfeld-conjecture-proof/index.json