Greg Egan on Nostr: This is certainly an interesting and important result in geometry, in and of itself! ...
This is certainly an interesting and important result in geometry, in and of itself!
Several professional mathematicians quoted here comment on the significance for the practice of mathematics of the fact that this was found by an LLM.
FWIW, I’m not hugely surprised that there are *some* interesting and important results that can be found this way — and once we make more use of formalisation there will be a whole lot more, with a lot less errors and slop to wade through — but I think it’s still likely that plenty of scope will remain for humans settling similar conjectures that software is unable to resolve, just as certain smart, tenacious humans have always found proofs that other smart, tenacious humans worked on for decades and were unable to resolve.
https://openai.com/index/model-disproves-discrete-geometry-conjecture/Published at
2026-05-20 23:22:01 UTCEvent JSON
{
"id": "2e77a71127d3e946b24ae1ebbcc780ea2954cd1c6c2f2415affe1c189964ec49",
"pubkey": "563821004c3e4e4f4fbef74a2657db3299a0e71a78214bf97db0b6ca05fb73d3",
"created_at": 1779319321,
"kind": 1,
"tags": [
[
"proxy",
"https://mathstodon.xyz/@gregeganSF/116609471073336438",
"web"
],
[
"proxy",
"https://mathstodon.xyz/users/gregeganSF/statuses/116609471073336438",
"activitypub"
],
[
"L",
"pink.momostr"
],
[
"l",
"pink.momostr.activitypub:https://mathstodon.xyz/users/gregeganSF/statuses/116609471073336438",
"pink.momostr"
],
[
"-"
]
],
"content": "This is certainly an interesting and important result in geometry, in and of itself!\n\nSeveral professional mathematicians quoted here comment on the significance for the practice of mathematics of the fact that this was found by an LLM.\n\nFWIW, I’m not hugely surprised that there are *some* interesting and important results that can be found this way — and once we make more use of formalisation there will be a whole lot more, with a lot less errors and slop to wade through — but I think it’s still likely that plenty of scope will remain for humans settling similar conjectures that software is unable to resolve, just as certain smart, tenacious humans have always found proofs that other smart, tenacious humans worked on for decades and were unable to resolve.\n\nhttps://openai.com/index/model-disproves-discrete-geometry-conjecture/",
"sig": "63542c506fb2bc75165627cc854023c005fa0e921fcd4d6370fcdc7a6111de65c5dfd33610fea122e1ebf31d4e6c2ad401effcda13dd728e5b74afd5d69a08f4"
}
