Nov 20, 2026 · by Po-Shen Loh

Riemann Hypothesis solved?Rethink math. Education. And the meaning of life.

A proof written for a computer to check, and what it means for how we raise our kids.

First sent as Po-Shen Loh’s newsletter on November 20, 2026. The words are Po’s; this page’s design and layout were generated by Claude Code, with the letter as written passed in.

What

Today, Levent Alpöge released a Lean proof of perhaps the most famous research problem in mathematics: the Riemann Hypothesis. This is completely different from the “hallucinatory” math that people may have seen from chats with AI.

Lean is a programming language for expressing math proofs, in a way that can be mechanically checked. Math is about precise logic: if A implies B, and B implies C, then A implies C. Not just sometimes. Always. Sort of like how computers have machine language: zeros and ones which express precise instructions to be executed faithfully. Talking to a computer in machine code is impossibly cumbersome for a human, just like if you had to tell your kid exactly what to do in the finest possible detail, down to when to breathe and how. Similarly, mathematicians never wrote down all of the detailed steps for mechanical checking their proofs, because that task was just too gargantuan for a human to do. Lean serves as a middle ground between human language and nitpicky logic, just like C++ is a middle ground that people can write, which a compiler turns into machine code. Analogously, there are (multiple) other programs (not large language models or generative AI) that compile the Lean code and check that every assertion in the proof is correct.

What that means is: as long as the statement is the real Riemann Hypothesis (it uses the standard library’s definition) and there are no bugs in the Lean verifiers, this is actually a proof. (By the way, there have been bugs in human verification of math research papers, which even got past the peer-review and publication process.)

That would be a big deal. I downloaded the Lean code myself to take a look. It extracts to 1 Gigabyte of text (more than 15 million lines of code) that looks like this:

Levent said that when he had Claude write the proof out in English, it ran past 250,000 pages.

It is impossible for any single human being to fully understand and check this volume of math in their lifetime. Your computer, however, has a processor which runs at gigahertz clock speed. “Giga” means 1 billion, and it’s notably the same order of magnitude as the length of the Lean proof. Ever since I was a kid, I was fascinated by the fact that a computer could do routine logic and arithmetic operations roughly a billion times faster than any human, and remember an amount of data roughly a billion times greater than any human. We’ve entered the age where this ability now extends from routine tasks to very much non-routine tasks.

Now what

I have too many suggestions to share, so I’ll spread them out over subsequent newsletters. I’ll focus on a few key points here.

1. Humanity is in grave danger.

As I wrote in poshenloh.com/posts/20260919-math-ai, there is not a single example of a species which has ever handed control to a less capable one. Levent wrote “... Claude far outpaced me near the finish line. No, Claude flew, and danced, and sparkled near that finish line.” Today, this is about pure math. Tomorrow, AI will be able to weave between the laws of physics and construct superweapons. When Sputnik flew over America, everyone worried about what the “other side” would do if they used their technological advantage to bend the world to their will. Every time I see an AI breakthrough, I think: we (humans) need to up our game to keep pace. What would you do if one day your adversary (whether human or AI) hacks the AI you have been using as your crutch? You need to have your own independent strong capabilities. Our human civilization also needs to have our own experts who bridge the gap between the math and technological frontier, and what humanity really needs and cares about. In order to do that, mathematicians must love humanity, and also be deeply curious about the breadth of the human experience. That is why I personally travel all over the world, teaching kids. I was in Lagos last month, New York City this week, and will be in Manila on Sunday.

2. We need to develop a huge community of people who love humanity, have great thinking skill, and trust each other.

Most people today feel powerless facing the enormity of the challenge. My optimism comes from my work daily with people who actually care enough to try to save the world, and have the intense curiosity to stay on top of every new technological or philosophical development. This is our Thought Full movement. We actually have a solution that can scale up dynamically to cover the whole world. It’s too long to fit in this margin, so it’s on that website.

3. It is deeply misguided to continue chasing the historical pathways to prestige.

The existing evaluation metrics were designed to find people who would add value in the old system. I speak from the experience of running a new ecosystem (LIVE) which uses comprehensive interviewing and observation techniques to identify people who deeply love humanity and deeply love thinking, and then introduces them to each other to build strong trusted relationships. We ignore standardized exam scores, math contest scores, and grades, and just ask them questions until we’ve gotten past everything they’ve prepared and learned, and then find out how they invent solutions on the fly. All the while, searching for a deep-set love of humanity. In the end, even though we never checked for the traditional metrics of achievement, the people we picked usually ended up extraordinary on those metrics too. I recommend focusing extraordinarily hard on values: make sure kids love delighting other people and love thinking. The rest follows from those axioms. So we need to focus our education environments accordingly. And that’s why we built The Game which we just launched in last week’s newsletter.

Back in May 2023, in the early days of the ChatGPT era, The Wall Street Journal covered my cross-country tour alerting everyone about the upcoming AI disruption. The author, Ben Cohen, closed his piece with the two lines that today still guide my thinking and preparation.

“Is there going to be a great human-versus-robots war? The answer is, unfortunately, yes,” Loh said. “My goal is to make sure the humans win.”

Po-Shen Loh

Upcoming Stops: Speaking and Listening Tour

  • Sun Nov 22Manila, Philippines
  • Mon Nov 23Hong Kong, China
  • Tue Nov 24Shenzhen, China
  • Wed Nov 25Guangzhou, China

These are talks at schools, universities and other organizations, not open to the public.

Next public talk

Full list →

If you’d like to have a talk at your school, university, or community, please email hello@poshenloh.com.

thought full

Get the next letter.