@semanticsam: Claude formally verified Fermat's last theorem in Lean 4 #math #ai #claude #lean4 Anthropic’s Claude has formally verified Fermat’s Last Theorem in Lean 4. The theorem challenged mathematicians for 358 years before Andrew Wiles solved it in 1995; Claude completed its formalization in 11 days. The proof contains more than 13 million lines, but Lean reduces the trusted surface to its roughly 8,000-line kernel and about 70 lines stating the theorem. This is how theorem provers can check proofs far too large for any person to read. As AI-generated proofs grow, what should mathematics prioritize: reaching the right answer or understanding why it’s true?
Literally who says “you can’t prove a negative”?????
2026-09-07 05:09:24
18
DW :
Solving something that has already been solved: it is just computing a theorem. That is what computers are supposed to do. Next year 11 days will seem like an ancient number.
2026-09-08 19:07:37
0
Drew :
Fermat, short for Fermatthew? 🤔
2026-09-08 17:09:45
1
Stupid cheesy mf :
andrew wiles proved this in 1994
2026-09-07 08:35:25
11
le_paz :
kinda old
2026-09-08 17:20:37
0
7Seven :
a proof is a proof
2026-09-08 14:38:28
0
Enchantbot :
I’d include the solution but I’ve run out of tokens
2026-09-07 20:46:00
10
jondoe :
are you IA?
2026-09-08 12:12:17
2
slea 🇦🇺 :
13 million lines definitely can't fit in a margin
2026-09-08 08:42:32
3
Dqffry :
i think the framing youve used deserves some scrutiny, this was an existing theorem that just got formalized in lean. the main constraint was building up the surrounding codebase to the proof, and according to what buzzard said on the xena project website, it sounds like what anthropic has horked up is more of a shortcut than a clear advancement. failing that, its just not that amazing since lean at large isnt how research is done due to the interfacial intertia that comes from the code
2026-09-07 03:10:21
13
Jexposition :
Noones been trying to solve this in lean lol it was solved.
2026-09-07 02:19:51
6
NeverSayDie❌🆘🇺🇸🇮🇱 :
maybe. The amount of mathematical machinery that was created in the pursuit of Fermat's last theorem was hardly basic mathematics. To say it requires little more than basic mathematics to understand is deeply misleading.
2026-09-08 01:52:36
1
user063495081 :
I think it is a waste of time unless it is useful. Can we have it focus on sometime like warp technology so we are not stuck on this planet forever?
2026-09-07 04:37:26
2
Untitled Cat Person :
It solved it, why humans (that couldn’t understand) need to understand it? Like, I think things like this will become more and more common from now on, it’s why we invented AI, to go further, right?
2026-09-08 10:49:33
1
davidanalyst :
when you say Fer mats last theorem, that says it all right there
2026-09-07 03:39:42
8
Tykkylumi :
This is what we need ai for, complex calculations and automatisation of tedious tasks, not the entertainment industry
2026-09-07 12:30:45
3
Jeremy Kun :
nobody is freaking out...
2026-09-07 15:26:16
1
Chemical chemical :
We are so close to amazing. If the fucknuts would just let us get there.
2026-09-07 01:21:26
1
Lorbmick :
the answer is both.
2026-09-07 13:09:01
1
KNfanacct :
Has Claude come up with a Unified Field Theory yet? I think we are all expecting it soon.
2026-09-06 23:27:23
5
Mark T 🇨🇦 :
Simplifying the proof is a different problem, probably a more approachable problem.
2026-09-08 12:40:36
0
To see more videos from user @semanticsam, please go to the Tikwm
homepage.