What OpenAI actually claimed
Crypto Twitter did what Crypto Twitter does, but the ETH in OpenAI's post is not Ethereum. It is the Exponential Time Hypothesis from 1999, proposed by Russell Impagliazzo, Ramamohan Paturi, and Francis Zane, which asserts that a classic logic problem called 3-SAT cannot be solved much faster than exhaustive search. OpenAI says its AI found a way around that assumption, and the originators are now examining the claim.
In plain English, 3-SAT asks for true or false values that satisfy many three-variable rules at once. OpenAI's paper presents a deterministic (no-randomness) algorithm that enumerates all satisfying assignments in subexponential time, i.e., its running time increases at a rate below every exponential function of the variable count. Because counting all solutions is tougher than merely checking if a solution exists, the deciding version would benefit too.
OpenAI says the system uncovered the refutation on October 16, and the result was mechanically verified in Lean on October 17. Both the write-up and the formal proof are on GitHub, the same channel as 722 math manuscripts OpenAI posted earlier this month, of which 162 included a computer-checked main result. The blog framed the stakes this way: "The refutation of the ETH reopens questions about how much faster these tasks could become," and may "help guide the search for better algorithms."
What ETH is and why researchers care
Hundreds of results lean on ETH to motivate why many challenges in geometry, databases, computational biology, and graph theory are broadly assumed to resist major performance gains. So if ETH falls, the speed limits in those areas might shift. OpenAI also emphasized limits here: the paper itself notes it does not claim that all NP problems are solvable in subexponential time, which means the famous P versus NP question is unaffected.
Practical payoffs are not on the doorstep. As OpenAI put it, "Finding such algorithms, and turning theoretical advances into practical improvements, will require significant further research."
Smart investors welcome new tools that can gently sharpen their decision making. We're hosting our first ever investor livestream on 11/17 to show you how to use ChatGPT and Claude to invest like a pro, hosted by Briefs Finance CEO Jaspreet Singh. Save your seat.
How scrutiny is being organized
Impagliazzo, Paturi, and MIT's Ryan Williams are leading the outside review, and they were explicit about boundaries: "Our involvement should not be taken as certification of the claim," they wrote, adding that "a result of this significance requires careful, independent scrutiny and an understanding that others can reconstruct, question, and build upon."
OpenAI said contributors to the effort signed no non-disclosure agreements and were unpaid. The company agreed to release all raw data and to fund a conference on the claimed result. Separately, a recent $5 million OpenAI donation to the Simons Foundation will fund community efforts - via workshops, conferences, and targeted initiatives - to assess and build on AI-generated mathematics.
The competitive backdrop and early market color
Anthropic quickly posted a repository with a paper dated October 26 titled "A 2^o(n) Time Algorithm for 3-SAT," crediting an internal research Claude model for the discovery. The repo underscores limits too, saying the finding is "purely asymptotic and says nothing about solving formulas in practice."
This sprint follows other formalization milestones. Claude helped formalize Fermat's Last Theorem in Lean in September, and Terence Tao observed the broader dynamic: "We have now seen that even the rumor of someone working on a problem can trigger a massive amount of AI-powered effort to flatten it before the original research project has time to reach its full potential."
Elsewhere on the screen, Bitcoin's snapshot showed BTC · USD at 81,078 dollars, down 4.24 percent, with a 24H high of 81,636 dollars, a 24H low of 80,334 dollars, and volume at 1.1 billion dollars.
What it could mean for your portfolio
If this refutation holds, it nudges the frontier on what might eventually be optimized in fields that lean on ETH-based limits. But the bottom line is timing: OpenAI and Anthropic both stress theory today, not immediate speedups in production code. For now, the investable signal is about the pace of AI-for-math research, the transparency of the verification process, and which platforms or research hubs become the place where new algorithmic ideas get vetted and, eventually, productized.
Learning how to use fresh technology thoughtfully can give investors a steady advantage. On 11/17, our CEO Jaspreet Singh goes live for our first ever investor livestream on how to use ChatGPT and Claude to invest like a pro. Register to join him.
