# Why do math when AI can?

> Math is facing an existential crisis due to advancements in AI, but the a16z Frontier research team is more optimistic than ever about the future of the field. If AI solved a problem you’d spent years working on, would you feel like celebrating? I’ve been turning that question over since July, when ...

- URL: https://a16zfrontier.com/posts/article/ai-math-human-why/
- Authors: [Robert Hackett](https://a16zfrontier.com/team/robert-hackett/)
- Published: 2026-11-21
- Focus areas: tech trends
- Tags: research, AI, AI Math

---

*Math is facing an existential crisis due to advancements in AI, but the a16z Frontier research team is more optimistic than ever about the future of the field.*

If AI solved a problem you’d spent years working on, would you feel like celebrating?

I’ve been turning that question over since July, when I read mathematician Kirwin Hampshire’s “[The Dark Night of Mathematics](https://kirwinhampshire.substack.com/p/the-dark-night-of-mathematics).” Hampshire had a picture of the life he was training for. As AI models began finding counterexamples to longstanding conjectures, he wondered whether he would ever get his shot at a discovery of his own. “I am suffering a profound spiritual crisis,” he wrote. “I have been screaming internally for days. It feels as though I am living inside of a nightmare.”

And that was before OpenAI [released hundreds of mathematical manuscripts](https://github.com/openai/math) in October. Members of the research community have expressed their [shock, excitement, and bewilderment since](https://proofsandprompts.com/2026/10/08/100-reactions-to-100-solutions/), as they contemplate what it means and how it feels to be doing math research in an age of accelerating AI.

One of the results in the OpenAI batch offered a [proof](https://github.com/openai/math/blob/main/preprints/Thompsons-group-F-is-nonamenable-September-23-2026/paper.pdf) settling a question the Cornell mathematician Justin Moore had returned to for 19 years. In 2010, Moore had bought a bottle of XO cognac outside Marseille to share with the others who had worked on the problem, whenever the answer might come. He still planned to open it, he told his colleague [Steven Strogatz](https://www.theatlantic.com/ideas/2026/10/ai-math-human-future/689133/), once he and his peers understood the proof and agreed it was right. But the answer had arrived among hundreds of others, “presented like severed rhino heads,” as Strogatz put it. Moore expected the toast to feel “more like a wake than a celebration.”

Our researchers at a16z Frontier know what it’s like to work on a problem a model might solve. They make extensive use of AI across cryptography, theoretical computer science, economics, and math, prying into stubborn old problems, testing ideas, developing proofs. We got them together to [compare notes — and one question kept coming up:](https://www.youtube.com/watch?v=KPIY2W4mphI) *What happens when proofs arrive faster than we can understand them?*

Dan Boneh, our senior research adviser and the head of Stanford’s applied cryptography group, offers a [thought experiment](https://www.youtube.com/watch?v=KPIY2W4mphI&t=1905s): Suppose a model proves that [P does not equal NP](https://www.claymath.org/millennium/p-vs-np/), a Millennium Prize problem that is considered one of the most important in all of computer science. (If true, this would mean, effectively, that there are problems whose proposed solutions can be checked efficiently even though no algorithm can solve every instance efficiently.) Now imagine the proof runs to a billion pages. Could any one person ever understand it? How long would it take to absorb, to explain, to teach, assuming that’s even possible?

Two days after our conversation aired, mathematician Levent Alpöge [announced](https://alpo.ge/remark.pdf) that he believes he has proved the Riemann hypothesis with the help of AI. The 1859 conjecture about the distribution of prime numbers is another one of the Millennium Prize Problems, with hundreds of results in number theory assuming it’s true. Alpöge says in the announcement: “It is natural to feel that this is an end. In my case a loss of a principal guiding light, but what I saw during the course of progress so amazed me that I believe it is one dawn across science of something far greater.”

The question of verification — what it takes to check others’ work — is one we spend a lot of time thinking about. Alpöge released his proposed proof in Lean, a programming language that lets computers check each logical step. But even if you could check every step of Dan’s billion-page proof, you’d still have the task of making it all intelligible — a task he says would take an “army of mathematicians and computer scientists,” over a very long period of time. (Good job security, at the very least.)

a16z Frontier researcher and computer scientist [Ittai Abraham](https://a16zfrontier.com/team/ittai-abraham/) (also behind the Decentralized Thoughts blog) thinks of mathematics as having two frontiers. The first marks the extent of discovery. The second marks how well humans understand what’s been discovered. 

Abraham points to a [1994 essay](https://arxiv.org/pdf/math/9404236) by mathematician William Thurston, which argues that mathematics is a social pursuit whose progress cannot be measured just by the sheer quantity of theorems. “What we are producing is human understanding,” Thurston writes. In his view, explaining ideas, challenging arguments, and helping others understand are central to the pursuit of mathematics.

That was Moore’s condition for opening the cognac: understanding the proof in company. A week after OpenAI’s release, Sang-hyun Kim and Yash Lodha posted [a simplified account of the same proof](https://arxiv.org/html/2610.18360v1). Their paper claimed no new discovery, but it gave other mathematicians a clearer explanation of why the result was true and how its reasoning could be used elsewhere.

[Scott Duke Kominers](https://a16zfrontier.com/team/scott-kominers/), our a16z Frontier research colleague, professor at Harvard Business School, and avowed math (and [puzzles](https://a16zfrontier.com/posts/podcast/puzzles-crypto-web3-with-a16z-podcast/)!) fanatic, knows the pleasure Moore was saving that cognac for. Kominers says [AI helped him solve a problem](https://www.youtube.com/watch?v=KPIY2W4mphI&t=3489s) he had been returning to since about 2007. When he wrote about the result to mathematicians whose work he had followed for years but whom he had never met, they wrote back delighted. In Scott’s telling, no one’s thrill was diminished by the use of any tools to get there. Scott now imagines researchers using AI to build a whole body of mathematics. (He also [wrote about this trend](https://a16zfrontier.com/posts/article/trends-ai-agents-automation-crypto/) early last year.) And as the tools become more powerful, he argues, our ambitions should grow with them.

The mathematician Frank Vallentin describes the pleasure of such a collaboration in the AI disclosure at the back of a [recent paper](https://arxiv.org/html/2609.35027v1). For years he carried a worn photocopy of a paper related to a problem he had loved since he was a student. Picture a repeating pattern of points called a lattice, with a ball of the same radius centered at every point. Do the balls cover the whole space, or is there a gap somewhere? Given the lattice and the radius, Vallentin wanted to know how hard it is for a computer to answer that question as the number of dimensions grows.

One Sunday in September, an AI model found an argument for a different problem about lattices. Vallentin wondered whether the same argument might work on his dear old question. After some back-and-forth, he found that it did. The resulting proof pinned down exactly how hard the covering problem is. Vallentin simplified the proof, found it beautiful, and took responsibility for its correctness, although he was unsure how to divide the credit. He was happy, however, just to be “the messenger of the proof,” he wrote.

So what does this mean for someone setting out to become a mathematician? Will there be anything left for them to do? 

Our researchers see plenty of work ahead absorbing, refining, and extending new results, explaining them to others, and advancing human understanding. Boneh has a wonderful metaphor for this: He [imagines AI as a huge tree](https://www.youtube.com/watch?v=KPIY2W4mphI&t=3706s), dropping succulent fruit for us to examine and enjoy. 

For someone hoping to make discoveries of their own, this may sound like cold comfort. But these tools also offer a chance to attempt work that would once have been out of reach — to build whole new bodies of mathematics, as Kominers imagines. Or as Alpöge writes, “I hope the argument provides a glimpse towards the miracles that lie ahead, and shows us the joy and wonder we are about to find in the simplest things.”

The people working and studying now will also get to decide which questions are worth all the effort in the first place. [Justin Thaler](https://a16zfrontier.com/team/justin-thaler/), a16z Frontier research partner at a16z Frontier and an associate professor in the Department of Computer Science at Georgetown University, who teaches computer science at Georgetown, points out that people will still want to make judgment calls of their own, however capable machines become. To make decisions well, you have to know the math and understand its consequences, especially when building systems other people will trust. In cryptography, judgments about which problems will remain hard to solve are already influencing [plans to secure the internet against quantum computers](https://csrc.nist.gov/pubs/ir/8547/final).

Mathematicians have spent a long time trying to hit difficult targets. There will be many more swiftly moving targets at far greater distances to come. Let’s welcome better aim.

\*\*\*

**Robert Hackett** is features editor and head of special projects at a16z Frontier.

\*\*\*

*The views expressed here are those of the individual AH Capital Management, L.L.C. (“a16z”) personnel quoted and are not the views of a16z or its affiliates. Certain information contained in here has been obtained from third-party sources, including from portfolio companies of funds managed by a16z. While taken from sources believed to be reliable, a16z has not independently verified such information and makes no representations about the current or enduring accuracy of the information or its appropriateness for a given situation. In addition, this content may include third-party advertisements; a16z has not reviewed such advertisements and does not endorse any advertising content contained therein.*

*This content is provided for informational purposes only, and should not be relied upon as legal, business, investment, or tax advice. You should consult your own advisers as to those matters. References to any securities or digital assets are for illustrative purposes only, and do not constitute an investment recommendation or offer to provide investment advisory services. Furthermore, this content is not directed at nor intended for use by any investors or prospective investors, and may not under any circumstances be relied upon when making a decision to invest in any fund managed by a16z. (An offering to invest in an a16z fund will be made only by the private placement memorandum, subscription agreement, and other relevant documentation of any such fund and should be read in their entirety.) Any investments or portfolio companies mentioned, referred to, or described are not representative of all investments in vehicles managed by a16z, and there can be no assurance that the investments will be profitable or that other investments made in the future will have similar characteristics or results. A list of investments made by funds managed by Andreessen Horowitz (excluding investments for which the issuer has not provided permission for a16z to disclose publicly as well as unannounced investments in publicly traded digital assets) is available at https://a16z.com/investments/.*

*The content speaks only as of the date indicated. Any projections, estimates, forecasts, targets, prospects, and/or opinions expressed in these materials are subject to change without notice and may differ or be contrary to opinions expressed by others. Please see https://a16z.com/disclosures for additional important information.*

---

Source: [Why do math when AI can?](https://a16zfrontier.com/posts/article/ai-math-human-why/) — a16z Frontier. See https://a16zfrontier.com/llms.txt for a site index.
