# OpenAI reports a subexponential algorithm for 3-SAT Markdown copy of https://realalpha.app/s/openai-reports-a-subexponential-algorithm-for-3-sat/ on Real Alpha, the best minutes from long interviews with the people building AI. Cite and link the page, not this copy. - Speaker: [David Friedberg](https://realalpha.app/people/david-friedberg/) (Host) · CEO, Ohalo · Founded The Climate Corporation, sold to Monsanto for about $1.1B in 2013. - Host: [Chamath Palihapitiya](https://realalpha.app/people/chamath-palihapitiya/) · Founder & CEO, Social Capital · Led Facebook's growth team; his firm backed Slack and AI chipmaker Groq. - Host: [Jason Calacanis](https://realalpha.app/people/jason-calacanis/) · Founder & CEO, LAUNCH · Early Uber angel; sold Weblogs Inc. to AOL; hosts This Week in Startups. - Show: All-In - Episode: [Midterm Blood Bath, $4T AI Sell Off, Data Center Freeze, What’s Next for Trump](https://realalpha.app/v/2v8b062QzTI/), Nov 7, 2026 - Moment: 1:24:41 to 1:29:04 of the episode (4 min), playing on https://realalpha.app/s/openai-reports-a-subexponential-algorithm-for-3-sat/; the episode from there: https://realalpha.app/v/2v8b062QzTI/?t=5081 ## Quote > “The solution, it's being evaluated now. So, you know, it's not fully kind of **verified by third parties yet**.” > — [David Friedberg](https://realalpha.app/people/david-friedberg/) at [1:27:50](https://realalpha.app/v/2v8b062QzTI/?t=5270) · All-In ## Takeaways 1. The result concerns a particular class of 3-SAT problems. 2. A subexponential algorithm would challenge the Exponential Time Hypothesis. 3. Friedberg says independent verification is still pending. ## About this segment David Friedberg describes an OpenAI paper claiming a subexponential-time algorithm for 3-SAT, challenging the Exponential Time Hypothesis. He says the 175-page proof still needs independent verification. From “Midterm Blood Bath, $4T AI Sell Off, Data Center Freeze, What’s Next for Trump”, Nov 7, 2026. [Watch the full episode](https://realalpha.app/v/2v8b062QzTI/) ## Transcript Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there. [1:24:44](https://realalpha.app/v/2v8b062QzTI/?t=5084) Please inform our audience about this. The reason for doing this Science Corner on this OpenAI publication is just to highlight some of the benefit and value that comes out of AI. ai these are like pretty substantial breakthroughs that are happening all the math hypotheses they've published on this one's called the exponential time hypothesis and it starts with like the most fundamental question in computer science which is can a problem be solved with a computer and if so how long does it take and that tells you whether it's actually practically computable this is the key point of computer science we want to create these machines that can solve our problems. And there's a class of problems that are called p problems, where you can solve them in what's called polynomial time, the bigger the problem [1:25:30](https://realalpha.app/v/2v8b062QzTI/?t=5130) of a particular type, the longer it takes in polynomial time just means size cubed or size squared, it tells you how many seconds or minutes or hours it might take on a specific computer to solve that problem. And so the problem as it gets bigger, takes more time. And then there's what are called NP problems. These are problems where you can very quickly figure out if the answer is correct. So like, you know, it can take a long time to solve a big Sudoku puzzle. But you can check the Sudoku puzzle in a couple of seconds and see if it's right. That's an NP problem. [1:26:01](https://realalpha.app/v/2v8b062QzTI/?t=5161) And the key question in computer science has been, do P problems equal NP problems, meaning, can you actually solve any problem that you can quickly check in a short period of time? And if so, then everything is computable the entirety of the universe is computable and that's the big question in computer science this idea of this exponential time hypothesis which was proposed in 1999 by a couple of guys and they said for a specific type of problem called a three sat problem which we don't really need to get into the amount of time it takes is exponential depending on how big the problem is so the bigger the problem gets it's some you know number to the nth power so it scales exponentially in terms of how long it would take a computer to solve it, which means that they have demonstrated that there is a class of problems that can never be solved [1:26:55](https://realalpha.app/v/2v8b062QzTI/?t=5215) in a short period of time. And therefore, that entire class of problems means that P does not equal anything. So that's been their hypothesis since 1999. And what OpenAI did is they disproved that. And they just proved it in 175 page paper. And it comes down to this one little formula, which you can pull up on the screen here, Nick, which we don't need to spend a lot of time analyzing to understand. But effectively, this little formula right here is OpenAI's paper demonstrating that the outcome of the paper demonstrating that this 3SAT problem, this particular class of problems in computer science can be solved in sub exponential time, which means it will not take an exponential amount of time. So they disproved this big exponential time hypothesis that everyone thought was real. And what that means is that P could still equal NP, [1:27:49](https://realalpha.app/v/2v8b062QzTI/?t=5269) which means that theoretically, every problem in the universe can be solved with a computer. The solution, it's being evaluated now. So, you know, it's not fully kind of verified by third parties yet. People are still trying to... understand it. It's 175 pages with a very novel set of approaches that it's taken to try and get there. They did the whole thing end to end in less than 10 days. Some reports have argued they did it in a few hours. And so this was a deeply challenging question in computer science and has been for 27 years. Is this exponential time hypothesis real or not? And they just proved that it's not. And they did it in a few hours. So it may be esoteric for the majority of people listening to this, but it does a couple of things. Number one, it keeps in play the idea that everything in the universe can be analyzed and every problem theoretically could still be solved [1:28:38](https://realalpha.app/v/2v8b062QzTI/?t=5318) with a computer. And number two is it's amazing that OpenAI was able to do this in a few hours and deliver this sort of a result that mathematicians and computer scientists have been struggling over for 27 years. And this proof is so far holding. So, you know, this is amazing. And so, everyone in computer science, I think, was kind of shocked when this came out. Absolutely. Yeah. It was shocked in terms of how it was presented, how they solved it. [1:29:03](https://realalpha.app/v/2v8b062QzTI/?t=5343) It was very novel.