Oct 13, 2026 · 27 min · 10 segments
OpenAI Releases 722 Math Papers from Secret Internal Model.In Episode 95 of What the AI?!, Jeff Keltner and Annie Delgado examine OpenAI's massive drop of 722 research papers generated from a single…
Annie DelgadoHostJeff KeltnerHostAll right, well, let's start off with something I had no understanding of at all, which is OpenAI dropped seven hundred and twenty-two math papers last week from, from an unreleased internal model.
Uh, they say most of it from a single prompt, go out there, solve all the hardest problems in the world of math or something like that, I don't know.
Um, the boldest claim, uh, boldest one of the papers claims progress on one of the most famous problems in math, but they've only had time to have computers, not even humans, but computers check like a fifth of these papers, so they're putting...
There's lots of existential angst, I have to say, in the math community over the release of like seven hundred and twenty-two papers in one day from OpenAI.
Well, uh, some of them are not happy, and then other ones just have like various random takes, like the guy who's like, "Yeah, but you know, the one we talked about last week, you stole my idea." And then the other guy that's like, "This is great.
We have so much more content." But mostly not happy because mostly the mathematics community is like, "Where are your receipts?" [laughs] Like, "How did it solve the problem? What was the answer?" And the answer was, "Uh, we didn't really have enough time to review all that, so just here's all the outputs."
I, I guess some, one of the complaints I've heard is that lots of them are kind of technical de-- like, like I...
Uh, not, cheating is the wrong word, but the, the sense I get from it, I ca- I can't understand any of the math papers anyway, so I can't tell.
But it's like they're solving them the way-- not in a way that humans would, in very kind of intricate kind of like s- kind of coming in the side door to solve these problems as opposed to the front door.
And I guess the, the finding of the front door is where a lot of the, uh, interesting thought happens, and so they're kinda like, "It's like this little side path in to find a solution to these," but I don't really understand at all other than there does seem to be existential dread that like the, the job of mathematicians is over.
I think it actually ironically just got harder, which is crazy because it was already a pretty hard job.
But like, you know, now the job is translating like so many of us into like the job of checking these really sophisticated things that are brain-bending questions and figuring out if they're real or not.
But it turns out that like there's not enough humans that actually understand that to actually do that.
So from, from what I understand, it was like twenty-two percent of these papers were checked by a machine, but not by a human, and so it's like, "Here's a bunch of stuff.
All right, well, let's start off with something I had no understanding of at all, which is OpenAI dropped seven hundred and twenty-two math papers last week from, from an unreleased internal model.
Uh, they say most of it from a single prompt, go out there, solve all the hardest problems in the world of math or something like that, I don't know.
Um, the boldest claim, uh, boldest one of the papers claims progress on one of the most famous problems in math, but they've only had time to have computers, not even humans, but computers check like a fifth of these papers, so they're putting...
There's lots of existential angst, I have to say, in the math community over the release of like seven hundred and twenty-two papers in one day from OpenAI.
Well, uh, some of them are not happy, and then other ones just have like various random takes, like the guy who's like, "Yeah, but you know, the one we talked about last week, you stole my idea." And then the other guy that's like, "This is great.
We have so much more content." But mostly not happy because mostly the mathematics community is like, "Where are your receipts?" [laughs] Like, "How did it solve the problem? What was the answer?" And the answer was, "Uh, we didn't really have enough time to review all that, so just here's all the outputs."
I, I guess some, one of the complaints I've heard is that lots of them are kind of technical de-- like, like I...
Uh, not, cheating is the wrong word, but the, the sense I get from it, I ca- I can't understand any of the math papers anyway, so I can't tell.
But it's like they're solving them the way-- not in a way that humans would, in very kind of intricate kind of like s- kind of coming in the side door to solve these problems as opposed to the front door.
And I guess the, the finding of the front door is where a lot of the, uh, interesting thought happens, and so they're kinda like, "It's like this little side path in to find a solution to these," but I don't really understand at all other than there does seem to be existential dread that like the, the job of mathematicians is over.
I think it actually ironically just got harder, which is crazy because it was already a pretty hard job.
But like, you know, now the job is translating like so many of us into like the job of checking these really sophisticated things that are brain-bending questions and figuring out if they're real or not.
But it turns out that like there's not enough humans that actually understand that to actually do that.
So from, from what I understand, it was like twenty-two percent of these papers were checked by a machine, but not by a human, and so it's like, "Here's a bunch of stuff.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.