The word doing the most important work in the headline is also the easiest to overlook: quasi.
A recent Lex Fridman clip is titled “Quasi-Riemann hypothesis proved by AI — mathematician explains.” That is not the same announcement as “The Riemann Hypothesis has been proved.” The distinction is mathematical, not cosmetic.
It deserves to survive the journey from a mathematician’s explanation to a headline, and from a headline to public understanding.
The classical Riemann Hypothesis concerns the non-trivial zeros of the Riemann zeta function. It says that every one of them has real part exactly 1/2. They must all stand on the critical line.
The reported quasi-Riemann result concerns the actual zeta function, too. It is not merely a look-alike function invented for an easier demonstration. But it asks something weaker: can a fixed boundary below 1 be established, beyond which no zeros occur?
OpenAI’s September 30, 2026 manuscript claims such a boundary at 7/8. A second manuscript, dated October 5, presents the weaker boundary 11/12.
If the 7/8 claim is correct, the symmetry of the zeta zeros would confine the non-trivial zeros to a strip between 1/8 and 7/8.
A strip is not a line.
Zeros could still stand inside that strip without standing at 1/2. The quasi-Riemann claim therefore leaves open precisely what the full Riemann Hypothesis requires us to settle.
This is not a reason to dismiss the work. A fixed zero-free boundary of this kind would be substantial mathematical progress.
But substantial progress deserves an exact account of what it accomplishes. It does not become more important by being mistaken for something else.
Nor does 7/8 mean “87.5 percent of Riemann solved.” Mathematical proof is not a progress bar. The remaining demand is not simply a smaller quantity of the same work.
The question is whether every non-trivial zero lies on one exact line. A wider permitted region does not answer it.
In an October 22 account from Georgia Tech, mathematician Alexander Dunn said he believed the 11/12 manuscript was correct. He and other mathematicians were continuing to review its argument and sources. Earlier techniques developed by Dunn and Maksym Radziwiłł contributed to the work.
Dunn also supplied the necessary caution:
“But it is too early to say whether it will help resolve the full Riemann hypothesis.”
OpenAI attributes the manuscripts to an unreleased internal model, while noting that the zeta work departed from its standard procedure and that the 11/12 exposition was human-edited for readability.
As of October 29, the sources cited here establish preprint claims and serious specialist scrutiny. They do not establish completed journal peer review of both manuscripts. I am reporting those sources, not independently certifying the proofs.
For those of us following the relationship between mathematics, AI, and human understanding, this is already an extraordinary subject. There is no need to erase the boundary between an achievement and the larger question it approaches.
Credit the theorem actually proved. Identify the human and AI contributions accurately. Keep the verification status visible.
And keep the word quasi.
A quasi-Riemann is not the Riemann.
Mathematicians who say we are now experiencing the greatest opportunity to get this right are the ones to follow. This is not a tragedy but the greatest opportunity.
Darkness at the break of noon
Shadows even the silver spoon
The handmade blade, the child’s balloon
Eclipses both the sun and moon
To understand you know too soon
There is no sense in trying
Bob Dylan, It’s Alright Ma (I’m Only Bleeding)
Further reading
