Subtitle: This article explores what an AI’s autonomous mathematical proof — and the week’s other frontier AI signals — actually require of us as entrepreneurs, not just what they tell us about technology.
On May 20, 2026, an AI produced a 125-page proof that disproved a mathematical conjecture Paul Erdős posed in 1946.
Eighty years. Some of the most capable mathematical minds in the world spent those eight decades working on variations of that problem. The conjecture sat in the category of “hard, open questions in mathematics” — the kind that occasionally get solved and generate headlines, but mostly just sit there, resistant to human effort.
A machine closed it in a session.
I want to be honest about my reaction when I read this. It was not excitement. It was not fear. It was something quieter than either of those.
It was the feeling I get when I stand somewhere vast and suddenly understand how small I am.
Key Takeaways
- An OpenAI reasoning model autonomously disproved an 80-year-old mathematical conjecture, producing a 125-page proof that represents AI reasoning at a level that matches — and in this case exceeded — the sustained output of human mathematical expertise.
- Demis Hassabis declared at Google I/O 2026 that we are in the “foothills of the singularity” — a precise framing from a scientist who has spent his career tracking this trajectory.
- Anthropic committed $200 million with the Gates Foundation to deploy AI in global healthcare and education.
- The right response to this news is not awe that paralyzes or fear that makes you retreat. It is recalibration — honestly updating your ceiling, and then deciding what to build to it.
- For those of us with a faith perspective, the question is not whether this should make us nervous. It is what responsibility this level of tool carries for the humans using it.
What I Actually Did With That Feeling
I sat with it for a few minutes. Then I asked myself the question I try to ask whenever something disrupts my mental model: what does this actually require of me?
Not what does it mean for the technology industry. Not what does it mean for the future of work. Not what do AI skeptics say versus AI boosters. What does this specific piece of information require of me, personally, as an entrepreneur who uses these tools in my business and teaches others to do the same?
Here is where I landed.
The tools I am recommending to entrepreneurs — the same tools I use to run my own operations — are operating at a level that I have been significantly underselling. Not intentionally. I have simply been calibrating my descriptions of what AI can do to what feels believable to the people I am talking to.
The Erdős proof changed my calibration.
The Capability Signal You Are Probably Misreading
When most people read about AI solving a hard math problem, they process it as a technology story. They think about AI, the industry, the competitive landscape, what it means for mathematicians. They do not think about the tools in their own daily workflow.
That is a mistake.
The reasoning models that produced a 125-page proof of a discrete geometry conjecture are available to you. Not as a research preview, not as a limited beta. As a subscription you can purchase and use today. The same underlying reasoning capability that autonomously solved an 80-year-old problem is the capability you are using when you ask Claude or ChatGPT to help you think through a business challenge, analyze a contract, or work through a strategy question.
I am not saying your business problems are as hard as Erdős’s conjecture. I am saying that if you are using AI to clean up your emails and write social captions, you are employing a system capable of 125 pages of sustained rigorous reasoning to do work that requires 3 sentences.
That is not a criticism. We all start somewhere. But the awareness of what these tools can actually do is worth having.
What Demis Hassabis Said and Why It Matters
At Google I/O 2026, Demis Hassabis said we are “standing in the foothills of the singularity.”
I want to be clear about who is saying this. Demis Hassabis is not an AI hype merchant. He is a neuroscientist, a game developer who built AlphaGo, and the person who has spent more time thinking about what artificial intelligence actually is — and what it is capable of becoming — than perhaps anyone alive. He co-founded DeepMind in 2010. He has spent 16 years building the systems that have produced AlphaFold, AlphaEvolve, Gemini, and the tools that now power hundreds of millions of interactions per day.
When he says “foothills of the singularity,” he is not speculating. He is reporting a trajectory from the person who has been measuring it with data for a decade and a half.
I am not here to make claims about the singularity or what it means theologically or philosophically. What I do want to say is that the person in the best position to make that statement made it publicly, at a major event, in front of thousands of people. That is worth pausing on.
The Gates Foundation Signal
The third piece of this week’s news is the one I keep coming back to.
Anthropic committed $200 million with the Gates Foundation to deploy AI in global healthcare, life sciences, education, and economic mobility over four years. The Gates Foundation does not make casual commitments. Their history is built on decades of rigorous evidence about what works in global health — and a willingness to make large bets when the evidence supports the investment.
The fact that the Gates Foundation partnered with Anthropic tells you something about what the evidence is showing about AI’s capability to change outcomes in some of the hardest domains in the world: disease treatment, educational achievement, economic access.
For me, as someone who cares deeply about faith and the way God works through human effort, this is meaningful. The tools being built are not only being used to automate business workflows. They are being aimed at the places where human suffering is most concentrated and most resistant to change.
That expands my sense of what we are actually dealing with. And it expands my sense of responsibility about how I teach people to use these tools.
The Question This Requires Entrepreneurs to Answer
Here is where I want to be direct with you about the personal challenge I have been sitting with since this week’s news.
The Erdős proof, the singularity framing, the Gates Foundation investment — together, these signals describe a moment when the tools available to everyday entrepreneurs have crossed a threshold that most people have not fully registered.
If AI can autonomously produce a 125-page mathematical proof, it can do things in your business that you have been assuming require sustained human expertise: deep analysis, rigorous logic, extended reasoning over complex multi-step problems.
The question this requires you to answer is: am I using these tools at a level that matches what they can actually do?
Not “am I using AI” — most entrepreneurs can say yes to that now. The question is whether the depth of your use reflects the actual capability you have access to.
In my own business, I have been pushing this frontier consistently. But reading about the Erdős proof made me identify at least three categories of decision-making in my business where I am still applying human reasoning alone to problems that AI could meaningfully contribute to. I am going to change that.
I am not sharing that to position myself as ahead or behind. I am sharing it because the honest response to genuine capability evidence is recalibration, and I am in the middle of doing it.
A Note for Those With a Faith Perspective
If you share my faith, I want to speak to something specific.
I know that advanced AI raises real questions for people who believe in a God who is sovereign over creation and who care about human dignity. I do not dismiss those questions. I think they are worth asking carefully.
What I have concluded — not from a position of certainty, but from sustained thought and prayer — is that the right response to these tools is not fear or avoidance, but stewardship.
We have been given tools of remarkable power. The question we are answerable for is not whether they exist. It is what we do with them, who we use them to serve, and whether our use of them reflects the values and the calling we have been given.
Anthropic’s $200 million commitment to global health and education is one answer to that question. It is not the only answer, and it does not resolve every concern. But it is a meaningful demonstration that the people building these tools are asking the right questions about what they are for.
My call, as someone who teaches entrepreneurship and AI adoption, is to help the people I work with answer those questions for themselves — and to build businesses that use these tools in ways they would be willing to stand behind.
Recalibrating Your Ceiling
After a week like this one, here is what I would encourage you to do.
Step 1: Identify three things in your business you have assumed require human expertise and challenge that assumption. Pick three decisions, analyses, or workflows that you have been keeping entirely in the human lane. Test AI on them, seriously, with real data and real stakes. You may be surprised.
Step 2: Try one reasoning-intensive use case you have not attempted. Not writing assistance. Something that requires sustained analysis: reviewing a contract for risk, analyzing your customer retention data for patterns, working through a pricing strategy with scenario modeling. These are the categories where the Erdős proof tells you the ceiling is higher than you thought.
Step 3: Read or watch the announcement from the Gates Foundation partnership. Not for the AI angle, but for the outcomes angle. What are they trying to change? What problems are they bringing AI to? It is grounding context for what these tools are for.
Step 4: Ask yourself the honest question. On a scale of 1 to 10, how deeply are you actually using AI — relative to what you know it is capable of? If the answer is below 7, what is the specific thing you are avoiding, and why?
Frequently Asked Questions
Should AI capability milestones like this change how an entrepreneur uses AI day to day?
Yes. Every significant capability milestone is a signal to update your ceiling — the level of complexity you are willing to give AI. Most entrepreneurs expand their AI use in response to direct experience, not news. Making a deliberate practice of updating your ceiling in response to evidence is a faster path to deeper adoption.
Does the Erdős proof mean AI is now smarter than humans?
It means AI demonstrated capability in one specific, well-defined domain that exceeded what human experts achieved over 80 years of effort. AI is not “smarter than humans” in any general sense — it is extraordinarily capable at specific tasks when given the right context and clear parameters. The lesson is about providing those conditions, not about a general intelligence claim.
How do I reconcile rapid AI advancement with a faith that values human dignity?
I hold both seriously. I believe human dignity is not diminished by powerful tools — it is expressed through how we choose to use them. The question is not whether to engage with AI, but how to engage with it in ways that serve human flourishing rather than undermining it.
What should entrepreneurs take from Demis Hassabis’s singularity framing?
Take it as a trajectory report from the most credible source available, not as a prediction of a specific date or event. Plan your business strategy on the assumption that AI capability growth will continue to accelerate and that the tools available to you in 2028 will be materially more capable than what you have today.
Is it responsible to be using AI in my business given the uncertainty about where this technology is heading?
Yes — if you are using it thoughtfully. The uncertainty is real. Thoughtful use includes staying informed, building human judgment into your workflows, and being willing to recalibrate as the landscape changes. Avoidance is not a neutral position — it has its own costs.
The Close
An AI solved a problem that stumped the best mathematicians on earth for eighty years.
I have been thinking about that number all week. Eighty years. Generations of extraordinarily intelligent people working on it. And a machine closed it.
I do not know exactly what that means for the future. I do not think anyone does.
What I know is that I am accountable for what I do with the tools I have been given — in my business, in my teaching, and in the way I help other entrepreneurs engage with what is available to them.
This week reminded me not to underestimate that accountability.
I hope it does the same for you.
Jonathan Mast is an entrepreneur, speaker, and founder who has been teaching AI adoption for entrepreneurs since the early days. He writes from a perspective that holds faith and technology in serious conversation. He lives in the Midwest with his family and is always working on something he has not told anyone about yet.
