InsightsNov 21, 2026 · 5 min read

An AI Just Helped Crack Math's Hardest Problem. Here's Why That Should Change How You Think About AI

A mathematician at Anthropic named Levent Alpoge just posted something that stopped the math world cold. He says that working with Claude, he found a proof of the Riemann Hypothesis.

If you didn't study pure math, here's the short version. The Riemann Hypothesis has been open since 1859. It sits at the center of number theory and touches how prime numbers are distributed. Mathematicians have built entire careers trying to chip away at it. Entire branches of math assume it is true, without anyone being able to prove it. It is the kind of problem that separates "hard" from "nearly impossible."

Alpoge posted the proof in Lean, a language that lets a computer check every logical step line by line. He says the method also covers the Generalized Riemann Hypothesis, an even broader version of the problem. Independent mathematicians have already run the Lean code without finding errors in the formal logic. The open question now is whether the statement Lean checked actually matches the Riemann Hypothesis as mathematicians have always posed it. That verification is still underway. Anthropic itself has been cautious here. Earlier this year, the company said it did not expect its techniques to lead to a full proof. This new claim goes further than that.

Whether or not the final verification holds up, the signal is already loud. A frontier AI model helped push on a problem that has resisted the best human minds for over a century and a half. That is not a chatbot trick. That is a model doing original mathematical reasoning alongside a trained expert, fast enough and rigorously enough to produce something the field has to take seriously.

Why this matters beyond math departments

You do not run a university math department. But the capability on display here is exactly the kind that shows up in your business, just pointed at different problems.

Think about what this proof actually required:

  • Holding a massive, structured problem in working memory
  • Testing approaches, discarding dead ends, and proposing new angles
  • Working alongside a human expert as a genuine thinking partner, not just a search engine
  • Producing output that could be checked, verified, and trusted line by line

That is the same shape of problem your finance team wrestles with during a complex reconciliation. It is the same shape as a legal team working through a tangled contract, or an operations leader trying to find the one bottleneck hiding inside a messy supply chain. The tools getting good enough to help solve a 167 year old math problem are the same tools that can sit inside your departments today.

What this means for how you roll AI out

A breakthrough like this is a reminder to large companies of three things.

First, the ceiling on what these models can reason through keeps moving. If you built your AI strategy around last year's capabilities, you are underestimating what is available to you right now.

Second, verification matters as much as the output. Notice that even with a stunning claim like this, the real world response was not blind trust. It was independent mathematicians checking the work line by line before anyone called it settled. That is exactly the discipline large companies need around AI output inside finance, legal, compliance, and IT. Trust the tool to do the heavy lifting. Build a process to check its work before it touches a customer, a contract, or a filing.

Third, the gap between what AI can do and what most companies actually use it for is still enormous. Most enterprise AI deployments today are chatbots answering simple questions. Meanwhile, the frontier of these models is tackling problems that have stumped experts for generations. That gap is where the real competitive advantage sits.

The takeaway

You don't need your teams proving theorems. You need them using this same level of reasoning power on the problems sitting in your own departments right now, with the right training and the right governance around it. The companies that figure out how to responsibly put frontier level AI reasoning to work inside finance, operations, legal, and customer service are going to pull ahead of the ones still treating AI like a search bar. The math world just got a preview of how far this technology can reach. Your business should be asking the same question about its own hardest problems.

RizeTech

Building AI departments inside enterprise companies

Get in touch

Talk about an AI Audit

A short first call, no pitch and no price. We learn how your company runs and where the hours go. If it fits, we scope the audit together. Book a time, or send a note and we reply within one business day.