Gemini did these, with a statement from the president of the IMO even confirming their results. I think the result is meaningful
2025-07-21 22:50:33
2
NEW TECH MATTERS :
OpenAI claims a “gold medal in math” for ChatGPT. Reality: can’t count letters in a word.
2025-07-22 01:25:18
11
Zaddycoin :
The funniest part is deepmind ratting them out and then posting the same thing after
2025-07-24 04:17:29
0
Heru-ur the Great :
They more concerned about funding than trust. And at this time particularly government funding.
2025-07-22 20:32:44
1
Bruce Fraser :
I think there is a disagreement here over what we mean by “intelligence.” If we are measuring AI against how humans solve problems, then we want LLMs to use human-like methods. If we measure intelligence by output, then we will be okay with “six LLMs in a trench coat.”
2025-07-21 20:00:43
4
itommalcolm :
super important, please keep at it 👍
2025-07-22 00:49:11
1
Somatic_Stretching :
Chay gpt is starting to teach me coding. and suddenly it's become very strange to me how AI developers push the idea of AGI, or even say AI would become dangerous to us --on its own. I don't mean if humans use ut badly. that is definitely a concern
2025-07-22 12:52:16
1
Umande Africa :
Very helpful. Regarding OpenAI, it was a downright lie, Nate. No need to sugar-coat it. One would've thought that they'd have learnt from Grok's debacle just several days ago and had their PR on point. How disappointing, because they needn't have engaged in such deceitful behaviour.😖
2025-07-21 22:24:21
3
l_ron_hubbard_official :
Like taking the SAT at your house 🤣
2025-07-23 01:25:55
0
depressedduck :
openai is like Tesla. they always make claims..
for now Google Gemini is the only ai with gold in math olympiad
2025-07-22 13:28:17
1
MickyMouse2 :
Trust is simpler if you just don’t listen to Altman and Musk
2025-07-22 07:29:44
7
johnny :
I study math and do undergrad competitions, and I also use AI quite frequently. I think i would agree that it’s very good at arriving at the right answer but the intermediate steps are shady. It’s kinda like a low resolution camera, the big picture looks ok but once you zoom in it becomes blurry.
2025-07-21 19:48:33
9
saint.null :
Isn’t that a thing in Maths? That you get graded on your working, not just the answers?
2025-07-22 04:05:14
1
Unsolicited Advice :
Nope, correct answers fast is the only point, thats all anyone cares about.
2025-07-21 21:08:37
1
user3776879128979 :
Scam Altman lie? Never!!! 😂😂
2025-07-21 20:17:46
17
NuTropics :
Results are results
2025-07-21 18:15:54
2
jcraig949 :
Claude keeps telling me I have production ready code that’s ready to ship, yet it doesn’t compile and mocked 2/3 of the code base
2025-07-21 19:06:48
25
nikomackay :
I was shocked because I catch simple math errors all the time for my chatGPT application 😂
2025-07-21 18:43:38
71
Neelay Chakravarthy :
Reminds me of the Majorana announcement in quantum computing from a while ago
2025-07-22 01:02:22
1
ak :
Altman like
2025-07-22 17:32:36
1
Jfp898 :
Someone taught ChatGPT to cheat. I think it might have been learning off me.
2025-07-22 09:51:40
1
andyeun :
Obviously for them the end result is important, not the details 😅😁 they are a company with a product and the sustained hype translates into money.
2025-07-22 20:49:03
2
etherealomega :
I’m always suspect of these claims primarily from OpenAI. They’ve gotten greedy and make a lot of half-truth claims. I’m looking at Meta’s proposed new datacenter the size of a small town and it’s becoming obvious that in order for them to scale to a size that will allow actually general adoption of agentic AI leveraging expert methodology we’re looking at insane power and raw material demands. I think in and of itself is a bit of a known thing that none of the major model makers want to have a real conversation about.
2025-07-21 19:22:15
4
Joshua :
There throwing off the gloves now to be first,right or wrong.
2025-07-21 18:11:30
8
osomarocks :
Nate, I recieved a tik Tok message from u, just making sure it is u
2025-07-22 17:26:56
1
To see more videos from user @nate.b.jones, please go to the Tikwm
homepage.