B
@baallives
GPT-5.4 Pro has cracked one of the Erdős problems that remained unsolved for over 60 years, using a 23-year-old.
12
0
406
@kimmonismusposted on X
A 23-year-old has cracked one of the Erdős problems that remained unsolved for over 60 years, using ChatGPT 5.4 Pro. Now image what ChatGPT 5.5 Pro will be capable of. And remarkably, it was done in a single pass. Total solve time: just 1 hour and 20 minutes.
View original tweet on X →
Community Sentiment Analysis
Real-time analysis of public opinion and engagement
Sentiment Distribution
83% Engaged
59% Positive
24% Negative
Positive
59%
Negative
24%
Neutral
18%
Key Takeaways
What the community is saying — both sides
Supporting
1
Prompting strategy matters:
framing the task as a test and telling the model not to search the internet reportedly made it more willing and capable to attempt the proof.
2
Human+AI collaboration
is emphasized — the result is portrayed as a joint effort (GPT-5.4 Pro working with a 23‑year‑old), not a pure solo machine breakthrough.
3
Speculative compute extrapolation:
commenters imagine near‑godlike capability if an AI had Dyson‑sphere scale power and millennia of thinking.
4
Predictive optimism:
it’s only a matter of time before people routinely use AI to discover scientific breakthroughs.
5
Model‑version hype:
some expect GPT‑5.5 to be “ridiculous” compared with current iterations.
6
Credit the present:
others push back, arguing 5.4 Pro is already doing the heavy lifting and deserves recognition now.
7
Workflow friction:
critics note a nontrivial chunk of time (about 20 minutes) was spent arguing over format, highlighting human–AI coordination costs.
8
Resource curiosity and concern:
people wonder how many tokens or how much compute the 1 hour 20 minute thought log consumed.
9
Hands‑on enthusiasm:
some replies are ready to line up unsolved problems and set their next‑gen chat model to tackle them one by one.
Opposing
1
Credit the human, not the version:
the solution was human-directed, and that distinction matters more than whether it was labeled 5.4 or 5.5.
2
Perceived regression in 5.5:
several users say GPT-5.5 feels lazier and “doesn’t think as hard” compared with 5.4, even under identical settings.
3
Not a general breakthrough:
benchmarks show strong performance on contest-style math (e.g., ~77%) but weak on open-ended research (~25%), so this case looks like a fortunate solution of “low-hanging fruit,” not evidence of new mathematical reasoning.
4
Practical skepticism and hype pushback:
critics call the example “slop” and warn against excitement until models deliver tangible, real-world value (e.g., “call me when it can make RAM cheaper”).
Top Reactions
Most popular replies, ranked by engagement
R
@RoenHistory
Honestly, for whatever reason, even though I'm on ChatGPT Pro, I find GPT-5.5 somewhat lazy, it doesn't think as hard on topics as 5.4, even with identical settings. I'm not sure why.
4
0
392
S
@ShreyasNalle
d, AI will surely serve drastically as a reason for humanities advancement. We all know everything rests on mathematics and biology. Just imagine what heights we can achieve if an AI can solve a problem which has unsolved since 60 years in just 80 mins. Absolutely insane
1
0
52
M
@msetechnologist
Imagine what a 100% Dyson sphere powered AI thinking for 10000 years will be capable of
1
0
64
H
@heygauravbhatia
the model didn't solve it. the human directed it. that distinction matters more than the version number.
1
1
119
F
@frog_omo
"Now imagine what 5.5 will do" is the wrong takeaway because GPT-5.2 scores 77% on competition mathematics but only 25% on open-ended research problems. This was one problem where the AI happened to try an unconventional approach that worked. It's not a general mathematical
0
0
126
Report an Issue
Found something wrong with this article? Let us know and we'll look into it.