Green's Problem 64 proved in Lean: infinitely many primes p for which p-2 has an odd number of prime factors
GPT-6 Astra, in a repository run by Igor Inozemtsev, produced an 83-page proof and a Lean formalization, registered on Palomar, that infinitely many primes p have p-2 with an odd number of prime factors counted with multiplicity, Problem 64 on Ben Green's list of open problems, drawing on an October 2026 manuscript by Jizhou Guo that the agent consulted.
- Lab
- Independent
- Model
- GPT-6 Astra (OpenAI)
- Field
- Mathematics
- Date
- 2026-11-13
- Human collaborators
- Igor Inozemtsev
What was found
Green's list notes that with the question is a weak form of the twin prime conjecture: the integers with an odd number of prime factors have density one half, so it asks for infinitely many primes in a set of density one half, but the parity problem of sieve theory blocks the usual methods. Chen proved in 1966 that has at most two prime factors infinitely often without deciding parity, and Pintz showed in 2010 that has an odd number of prime factors infinitely often for at least one among . The proof assigns weights to integers in with odd, sieves the values using a shifted correlation estimate, and bounds the contribution of products of two large primes from above, leaving positive weight on primes for every large . The Lean statement, written with Mathlib's , is identical to Green64.green_64 in Google DeepMind's Formal Conjectures, where it is categorised as an open research problem. The project's timeline records the proof accepted by the Lean kernel, by nanoda and by con-ron on 142,651 declarations, and the full Palomar workflow passing in CI on 13 November.
Novelty check
Green's list in its current form gives no update recording a solution of Problem 64, and Formal Conjectures marks it open. arXiv, 2026-11-16: abs:"prime factors" AND abs:"p-2" AND abs:odd returned 6 records, the only relevant one being Pintz's 2010 result for some shift up to 16 (1004.1065); abs:Liouville AND abs:"shifted primes" returned 2, neither on this question. The repository itself credits a manuscript by Jizhou Guo, Shifted primes with a prescribed parity of the number of prime factors, posted on GitHub in October 2026 and returning 404 by 12 November, which its agent consulted from 4 November to guide the final formalization, and two OpenAI manuscripts of September 2026 on prime predecessors with an even number of prime factors and on smooth shifted primes. It claims no priority. This entry therefore records the first machine-checked proof; the first claimed proof may be Guo's, which this registry does not carry and which was not read for this entry.
Caveats and known objections
Graded formal on the Palomar registration, trust level high, and the statement has an anchor written independently of the proof: it matches Formal Conjectures exactly. The two OpenAI manuscripts and Guo's are not independently verified, but the repository says every supporting input the formal statement needs is proved inside the Lean development, so the formal result does not rest on them. The 83-page article was reviewed only by AI models, and the README says no human mathematical review is claimed. Priority is not established, as set out under the novelty check. Autonomy is ai-led on the strictest defensible reading: proof development and formalization are credited to the model alone, and the human role beyond choosing the problem and running the project is not described.
Machine-checked elsewhere
PalomarPrimes p for which Omega(p - 2) is odd (Green's Problem 64)
PALOMAR-2026-11-13-000010 · inozemtsev/formal-math at 43f0988e15c2 · checked 2026-11-16
ProvedGreen64.green_64
A Lean proof that typechecks against the recorded statement at a pinned commit under a declared axiom set, checked mechanically rather than by human review. It does not certify that a result is new or of research interest: the only filter on that is a language-model screen, and Palomar states that it adds no human editorial step.
Trust level high. Version 3 of the record, with Green64.green_64 as the sole principal result; the project sits under proofs/green64.
No independent check is recorded on this entry yet.
Reading the primary source closely enough to say whether it supports the claim counts as a check, and you are credited on the entry.
Or on GitHub: submit a check challenge the grade send a correction or send a pull request
Flag this for triage
Signals order the review queue and nothing else. They are never published, and they never move a grade: that takes a citation.
Entry history (1 event)
- AddedEntered the registry graded Formally verified and AI-led.
Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.
Graded formally verified for verification and ai-led for autonomy. What these mean.
Cite this entry
whataifound.org. (2026). Green's Problem 64 proved in Lean: infinitely many primes p for which p-2 has an odd number of prime factors. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-11-13-green-problem-64
BibTeX
@misc{whataifound-independent-2026-64,
title = {Green's Problem 64 proved in Lean: infinitely many primes p for which p-2 has an odd number of prime factors},
author = {{whataifound.org}},
year = {2026},
howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
note = {Result by Independent. Verification: Formally verified. Autonomy: AI-led.},
url = {https://whataifound.org/finding/2026-11-13-green-problem-64}
}