Epoch AI has confirmed that OpenAI's GPT-5.4 Pro solved an open problem in Ramsey hypergraph theory — not a textbook exercise with a known answer, but an unsolved problem from the research frontier. The confirmation comes via Epoch's FrontierMath benchmark, which tracks AI performance against problems that professional mathematicians consider genuinely hard.
Let's be precise about what happened. FrontierMath isn't your typical benchmark of competition-math problems that a smart undergrad could grind through. It's a curated set of open and near-open problems spanning combinatorics, number theory, algebraic geometry, and more. When FrontierMath launched, frontier models were solving low single-digit percentages of the benchmark. The problems are designed so that even domain experts might spend days or weeks on them.
A Ramsey hypergraph problem sits in extremal combinatorics — the branch of math concerned with how large or small a structure can be while guaranteeing certain properties. Classical Ramsey theory asks: how big does a graph need to be before it must contain a monochromatic clique? The hypergraph generalization pushes this into higher dimensions, and the bounds are notoriously difficult to pin down. Many open problems in this area have resisted decades of human effort.
GPT-5.4 Pro producing a verified solution to one of these problems is qualitatively different from passing a coding interview or scoring well on GPQA. This is novel mathematical reasoning — generating a proof or construction that didn't previously exist in the training data, because it didn't exist anywhere.
The HN discussion (345 points) reflects genuine surprise even among the typically skeptical crowd. The key question isn't whether the result is correct — Epoch's verification process is rigorous — but what it implies about the trajectory. If models can solve open problems in extremal combinatorics today, the ceiling for AI-assisted mathematical research just moved significantly.
For practitioners, the immediate takeaway is narrow but important: the gap between 'AI can verify proofs' and 'AI can discover proofs' just closed for at least one non-trivial case. If you're in any field that touches combinatorial optimization — network design, constraint solving, scheduling — the research tools available to you are about to get meaningfully better.
The broader signal: we've moved past the phase where AI math benchmarks measure pattern matching on known problem types. We're now measuring genuine mathematical creativity. Whether that's a fluke or a trend, the next FrontierMath update will tell us.
I don't know why I am still perpetually shocked that the default assumption is that humans are somehow unique.It's this pervasive belief that underlies so much discussion around what it means to be intelligent. The null hypothesis goes out the window.People constantly make comments like &q
I have long said I am an AI doubter until AI could print out the answers to hard problems or ones requiring tons of innovation. Assuming this is verified to be correct (not by AI) then I just became a believer. I would like to see a few more AI inventions to know for sure, but wow, it really is a ne
For those, like me, who find the prompt itself of interest …> A full transcript of the original conversation with GPT-5.4 Pro can be found here [0] and GPT-5.4 Pro’s write-up from the end of that transcript can be found here [1].[0] https://epoch.ai/files/open-problems/gp
I like to imagine that the number of consumed tokens before a solution is found is a proxy for how difficult a problem is, and it looks like Opus 4.6 consumed around 250k tokens. That means that a tricky React refactor I did earlier today at work was about half as hard as an open problem in mathemat
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
I am kind of amazed at how many commenters respond to this result by confidently asserting that LLMs will never generate 'truly novel' ideas or problem solutions.> AI is a remixer; it remixes all known ideas together. It won't come up with new ideas> it's not because the mo