Remember that feeling in math class when the teacher wrote a problem on the board that made your brain hurt just looking at it? Well, imagine problems so hard that even the world’s smartest mathematicians have been stumped for decades. Now here’s the wild part: AI just started solving them.
Last week, I watched my nephew struggle with his algebra homework, and I couldn’t help but think about how OpenAI announced its next major model “Astra” by dropping solutions to ten previously unsolved mathematical problems. Not just any problems—these were questions that had mathematicians pulling their hair out for years.
Don’t worry if you’re thinking “math isn’t my thing.” This isn’t really about the math. It’s about what happens when AI starts doing things we thought only the most brilliant human minds could do—and what that means for anyone building with AI tools today.
The Day Mathematics Changed Forever
Let me paint you a picture of what actually happened. OpenAI didn’t just release another chatbot update. They dropped a 249-page paper showing their AI had solved mathematical problems that had been unsolved for over a decade. We’re talking about the kind of problems that win you the Fields Medal (basically the Nobel Prize of mathematics).
What makes this even more mind-blowing is the compute cost was only about $2,000. Think about that for a second. Problems that the brightest mathematical minds couldn’t crack after years of work were solved for less than the cost of a used laptop.
But before you start thinking AI has conquered all of mathematics, there’s an important detail: the AI couldn’t solve the really big ones—the Millennium Prize Problems. These are seven problems so important that there’s a million-dollar prize for solving each one, and only one has been solved since 2000.
You’ll see this pattern a lot with AI breakthroughs. It’s incredibly powerful at certain types of problems but still hits walls with others. Understanding this balance is crucial if you’re building with AI.
Why This Changes Everything (Even If You Never Touch Math)
Here’s where it gets interesting for those of us who aren’t mathematicians. The way AI approached these problems wasn’t just through brute force computation. According to the research, the AI completed “dimension-reduction style proofs and disproofs in multiple fields including geometry, algebra, and group theory”—basically, it found elegant shortcuts that human mathematicians had missed.
Think about what this means. If AI can find new ways to solve problems that experts have been working on for decades, what else might it help us see differently?
I’ve been experimenting with this myself. Last month, I was stuck on a coding problem—not a world-changing mathematical proof, just a tricky algorithm for sorting data. Instead of banging my head against the wall, I described the problem to Claude (an AI assistant) in plain English. Not only did it solve it, but it showed me an approach I’d never considered.
The pattern here is clear: AI isn’t just faster at doing what we already do. It’s finding entirely new paths we didn’t know existed.
The Human Side Nobody’s Talking About
There’s something Anthropic’s mathematician Levent Alpöge did that caught my attention. He used Claude (Anthropic’s AI) to disprove a mathematical conjecture that had stood for decades. But here’s the key: he didn’t just hit “generate” and walk away. He guided the AI, verified its work, and then had other mathematicians double-check everything.
This is exactly how we should be thinking about AI in our own work. You’re not being replaced—you’re being given a ridiculously powerful thinking partner.
I learned this the hard way when I first started coding with AI help. My initial instinct was to copy-paste whatever the AI suggested. Big mistake. The real magic happened when I started treating it like a collaboration: I’d explain what I wanted, the AI would suggest approaches, I’d test them, refine the questions, and iterate together.
Don’t worry if this feels overwhelming at first. Every expert using AI today started exactly where you are.
What Actually Works in Practice
Let me share what I’ve learned about working with AI on complex problems, whether you’re coding, designing, or even just trying to automate your daily tasks:
Start with clear problem definition. The mathematicians solving these problems didn’t just throw vague questions at the AI. They were specific about what they were trying to prove or disprove. When I’m stuck on code, I’ve found that spending five extra minutes clearly describing the problem saves hours of back-and-forth.
Use AI to explore multiple approaches. One fascinating aspect of the mathematical breakthroughs was that AI could simultaneously test different proof strategies. You can do the same thing. Ask for three different ways to solve your problem, then pick the one that makes the most sense for your situation.
Always verify and test. Just because AI solved those math problems doesn’t mean the solutions were immediately accepted. Independent mathematicians verified everything. In your work, this means testing code, double-checking facts, and using your judgment about whether the solution actually makes sense.
Build on what works. Once you find an AI approach that clicks for you, document it. I keep a simple text file of prompts that have worked well for different types of problems. You’ll see patterns emerge that you can reuse.
The Uncomfortable Questions We Need to Ask
There’s an elephant in the room we should address. If AI can solve problems that mathematicians spent careers on, what happens to expertise? What happens to the satisfaction of discovery?
OpenAI’s model still couldn’t crack the Millennium Prize Problems, which tells us something important: the hardest, most creative challenges still need human insight. But the bar for what counts as “uniquely human” keeps moving.
Here’s my take: instead of worrying about what AI might replace, focus on what it enables you to do that you couldn’t before. Before AI helped with coding, I could never have built the projects I’m working on now. The democratization isn’t about making experts obsolete—it’s about letting more people participate in creation.
Your Next Steps (Yes, You Can Do This)
If you’re feeling inspired but not sure where to start, here’s your homework for this week:
Pick one problem you’ve been putting off because it seems too complex. Maybe it’s automating a report at work, building that app idea you’ve been sitting on, or even just organizing your digital life better.
Now, instead of trying to solve it all yourself, open up Claude, ChatGPT, or whatever AI tool you have access to. Describe your problem in the plainest language possible. Don’t worry about sounding technical—remember, these AIs just solved math problems by finding patterns humans missed. They can handle your plain English just fine.
When you get a solution, don’t just implement it blindly. Ask “why does this work?” and “what if I changed this part?” You’ll see that the AI becomes more helpful the more you engage with it as a thinking partner rather than an answer machine.
The Real Magic Is Just Beginning
What excites me most about AI solving these mathematical problems isn’t the problems themselves—it’s what it represents. We’re entering an era where the biggest barrier to solving problems isn’t intelligence or expertise; it’s knowing how to ask the right questions and guide the tools we have.
Every time I see another breakthrough like this, I think about all the people who’ve been told they’re “not smart enough” or “not technical enough” to build things. Those barriers are crumbling, one solved problem at a time.
You don’t need to be a mathematician to benefit from what’s happening. You just need to be curious, willing to experiment, and ready to collaborate with AI rather than compete with it.
The mathematicians whose problems just got solved by AI aren’t becoming obsolete. They’re getting freed up to ask even bigger questions. The same opportunity is sitting in front of you right now.
So here’s my challenge to you: This week, pick one thing you’ve always wanted to build or solve but thought was too hard. Give it another shot, but this time with AI as your collaborator. You might be surprised by what you can achieve when you combine your creativity with AI’s problem-solving power.
Remember, just a few years ago, these mathematical problems seemed impossible. Now they’re solved. What impossible thing will you tackle next?