Post-Cutoff

Science & mathAalto University, Google DeepMind and Anthropic79 days after June 2026

ζ(5) proved irrational

Aabir Fauzan’s Zenodo preprint, the first such result since Apéry’s ζ(3) in 1978, is formally verified in Lean within a week, one formalization written by Claude

Result confirmed

Importance: historic (5 of 5)

The takeaway

On Sept 17, 2026 Aabir Fauzan (Aalto University) posted “ζ(5) is irrational” on Zenodo.

Status
Claim

Result confirmed

Our reporting
High confidence
Verification
Lean-verified
Importance
Historic (5 of 5)
Last verified
5 October 2026

Your AI and this story

  • GPT-6.2 Astra109 days after its cutoff
  • Claude Opus 5.579 days after its cutoff
  • Gemini 4 Hydrogen170 days after its cutoff
  • Grok 4.7109 days after its cutoff

None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 79 days before it.

Key facts

  • Preprint: A. Fauzan, ‘ζ(5) is irrational’, Zenodo 10.5281/zenodo.22826419, Sept 17, 2026; posted on Zenodo, not arXiv (about 16.8k views and 2.4k downloads by Oct 5). The record calls it the initial presentation, with a full manuscript and arXiv version to follow
  • Method (abstract): integer polynomials Q_n of degree 37n with 0 < Q_n(ζ(5)) < exp(−139n²/5), built as rationally normalized Hankel determinants with a positive moment representation and logarithmic-energy estimates. It also gives the irrationality measure bound |ζ(5) − a/b| > b^−260
  • Before this: Ball–Rivoal (infinitely many odd zeta values are irrational) and Zudilin (at least one of ζ(5), ζ(7), ζ(9), ζ(11) is irrational), but no specific odd value beyond ζ(3)
  • Paper’s AI disclosure: ‘A generative AI tool was used in a supporting role for editing the exposition, proofreading, LaTeX preparation, and consistency checks of the arguments, calculations, and references’
  • Frank Calegari (Sept 24): ‘The paper I saw gives me the impression of being almost entirely AI generated (if so, it then comes with a very dishonest disclosure), but never mind, thanks for the compute!’ He traces the key ideas to Prévost and Prévost–Rivoal
  • Lean formalization 1 (github.com/mo271/zeta5, Moritz Firsching, Google DeepMind; first commit Sept 23): complete with no sorry, axioms only propext, Classical.choice and Quot.sound. CI runs Comparator against Formal Conjectures’ RiemannZetaValues.irrational_five. It deviates from the paper in three local estimates (rank loss r, cutoff M = 400) but still proves irrationality. How it was produced is not stated
  • Lean formalization 2 (github.com/danromik/zeta5-irrationality): 70 files, about 29,400 lines, no sorry; the prime number theorem is assumed as an axiom. README: ‘The Lean code and the documentation were written by Claude (Anthropic), using the models Claude Opus 5 and Claude Opus 5.5, working as multi-agent workflows in Claude Code directed by Dan Romik’. Its pre-formalization audit found one inequality (p. 10) stated without proof and supplied a proof
  • Other efforts: an ‘autonomous’ formalization project (long-mathematics/zeta5-irrationality, Christopher D. Long) and a copy in Vilin97/lean-pool. One write-up reports the method handles ζ(k) for k = 2, 3, 4, 5 with changed parameters
Show 1 more
  • Reactions: Elliot Glazer (Sept 22, ~81k views): ‘zeta(5) has very likely been proven irrational … I had Astra study the argument, and it fully vouches for its correctness’. Alex Kontorovich (Sept 23, ~74k views): ‘WOW!! Zeta(5) is irrational! … Amazing what we’ll learn (with AI help)’

What happened

Apéry’s 1978 proof that ζ(3) is irrational was the last time a specific odd zeta value was shown to be irrational. Later work (Rivoal, Ball–Rivoal, Zudilin) proved only statements about families, such as “at least one of ζ(5), ζ(7), ζ(9), ζ(11) is irrational”. On Sept 17, 2026 Aabir Fauzan, giving an Aalto University address and with no earlier research papers, posted a 31-page proof for ζ(5) on Zenodo. Commentators guessed he had no arXiv endorsement.

Verification was unusually fast. Elliot Glazer publicised the paper on Sept 22. On Sept 23 Moritz Firsching (Google DeepMind, a maintainer of the Formal Conjectures benchmark) pushed a complete Lean 4 formalization, and Alex Kontorovich announced it. Comparator CI then confirmed that it proves exactly the benchmark statement irrational_five with only the standard axioms. A separate formalization by Dan Romik was, by its README, written entirely by Claude Opus 5 and Opus 5.5 agents in Claude Code. It was done without consulting the other formalization, and its audit filled one gap in the paper.

The AI role in finding the proof is unclear. The paper’s disclosure is limited to editing, LaTeX and consistency checks. Frank Calegari wrote that the paper looks “almost entirely AI generated” and, if so, the disclosure is “very dishonest”. He then used ChatGPT to trace where its ideas came from. Andreas Holmstrom classifies it as “Human + AI (supporting role)”. No AI system has been named as the author of the argument, and Fauzan has made no public statement that we found.

Why it matters

It is the most famous number-theory result of the AI-assisted 2026 wave: a problem experts had worked on for about 48 years. It was settled in a non-traditional venue and machine-verified in Lean within a week, with AI agents doing at least one formalization. That pattern (preprint, then Lean, then expert reading) reverses the usual peer-review order. The paper has not been refereed, and the ZFC-level guarantee rests on the Lean checks, one of which assumes the prime number theorem.

Sources

10 sources from 6 sites. Numbers match the chips in the text.

10 sources: 5 primary, 5 reactions

Primary

  1. Zenodo: A. Fauzan, ζ(5) is irrational (Sept 17, 2026)zenodo.org, paper
  2. Lean formalization (Moritz Firsching): mo271/zeta5github.com, code
  3. Lean formalization written by Claude (Dan Romik): danromik/zeta5-irrationalitygithub.com, code
  4. Formal Conjectures: RiemannZetaValues statementgithub.com, code
  5. long-mathematics/zeta5-irrationality (autonomous formalization attempt)github.com, code

Reactions

  1. Frank Calegari (Persiflage): zeta(5) is irrational (Sept 24)galoisrepresentations.org, discussion
  2. Andreas Holmstrom: Number theory breakthroughs at the dawn of a new erapeakmath.org, discussion
  3. Bogdan Grechuk: Zeta(5) is irrational (exposition, Sept 29)functor.network, discussion
  4. Elliot Glazer on X (Sept 22)x.com, discussion
  5. Alex Kontorovich on X (Sept 23)x.com, discussion

Changes

  • Filed (the dataset had missed it since Sept 17)

Status

Claim

Result confirmed

Our reporting
High confidence
Verification
Lean-verified
Importance
Historic (5 of 5)
Last verified
5 October 2026

Sources at a glance

10 sources: 5 primary, 5 reactions

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 6 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

    New record bound on the irrationality measure of ζ(2), μ ≤ 5.0495243, released as a 92k-line Lean proof written by Claude; beaten a week later by a Claude-assisted paper

    Result confirmed

Posts we archived

Videos

The whole topic