Watch cream swirl into coffee. Feel a plane shake in turbulence. Check tomorrow’s weather forecast. All three come from the same set of equations: the Navier-Stokes equations. They are the rulebook for how liquids and gases move.
For over 160 years, mathematicians couldn’t answer one basic question about these equations. Do they always make sense? Today, that question got a huge, and controversial, answer from an AI. Let’s walk through it simply, one step at a time.
Step 1: What Are the Navier-Stokes Equations?
Picture a tiny drop of water moving through a river. Three everyday forces push it around.
What pushes a bit of fluid around: pressure, viscosity, and momentum.
● Pressure. Water gets shoved from crowded areas toward empty ones, the same way air rushes out of a popped balloon. Think of rush hour traffic: cars in a packed lane keep pushing forward into any open space ahead. Fluid does the same thing.
Pressure works like traffic. It always pushes toward empty space.
● Viscosity. This is friction. Pour a glass of water and a jar of honey side by side and you’ll feel it immediately: the water splashes out in a second, while the honey creeps out slowly. That resistance to flowing is viscosity.
Water flows fast because it has low viscosity. Honey flows slow because it has high viscosity.
● Momentum. A moving drop tends to keep moving the same way, the same reason a rolling ball keeps rolling across a floor until something stops it.
The Navier-Stokes equations simply add these three everyday effects together. They’re named after Claude-Louis Navier, a French engineer who wrote an early version in 1822, and George Stokes, who tested and refined it. Two centuries later, we still use these exact equations to design airplanes, forecast hurricanes, and model blood flow. They work extremely well in practice.
Where You Already See This Every Day
You don’t need to be a mathematician to run into Navier-Stokes. It’s already working behind the scenes all around you.
The same equations describe air over a wing, storms in the sky, and blood moving through your body.
● Airplane wings. Engineers use these equations to predict how air bends around a wing to create lift.
● Weather forecasts. Meteorologists use them to model how storms and hurricanes swirl and move.
● Blood flow. Doctors use them to understand how blood moves through your arteries and veins.
Step 2: So What’s the Actual Problem?
Here’s the twist. Engineers can plug these equations into a computer and get answers that match the real world closely enough to build airplanes. But mathematicians want more than “close enough.” They want a guarantee.
Specifically, they want to know this: if you start with a normal, smooth flow of water, is it guaranteed to stay smooth forever? Or could it suddenly do something impossible, like reach infinite speed?
Ordinary flow stays at a normal speed. A “blow up” flow shoots to infinite speed at a specific moment.
Nothing in real life moves at infinite speed. So if the equations predict that it can happen, starting from totally ordinary conditions, that’s a red flag. Mathematicians call this a blow up. Nobody could prove whether it happens or not. This puzzle is nicknamed the Navier-Stokes existence and smoothness problem, and it’s so famous that in 2000 it was named one of seven “Millennium Prize Problems,” each with a $1 million reward for whoever solves it. Only one of the seven had ever been cracked before now.
Step 3: Why Was This So Hard?
The equations describe motion that constantly folds back on itself. Fast water pushes on slow water, which changes the pressure, which changes the speed again, over and over. This kind of self feeding loop makes the math incredibly tangled.
A simple comparison helps here. Smooth, orderly flow is easy to describe. Chaotic, swirling flow is where things get unpredictable, like the difference between a calm bathtub and a bathtub during a splash fight.
Smooth (“laminar”) flow versus chaotic (“turbulent”) flow.
Turbulence, the messy swirling kind of flow, is one of the last great unsolved problems in physics. Nobody fully understands why it happens or how to predict it in detail. The blow up question sits right at the heart of that mystery. Is there a point where the equations simply break down?
Mathematicians had chipped away at parts of the puzzle for decades. In 1934, a French mathematician named Jean Leray proved that “loose” versions of solutions always exist, but he couldn’t show they stayed smooth or that there was only one possible answer. Other researchers proved things about simpler, two dimensional versions of the problem. The full three dimensional question stayed open.
Step 4: An AI Cracks It, Maybe
On September 8, 2026, OpenAI announced that one of its AI systems had produced a proof. According to the company, the proof shows that it is possible for a perfectly ordinary, smooth flow of fluid to reach infinite speed in a finite amount of time. In other words, blow up can happen.
If that holds up, it means the Navier-Stokes equations, despite being incredibly useful for everyday engineering, aren’t a perfect mirror of reality in every mathematical scenario.
OpenAI didn’t just take the AI’s word for it. The proof was checked using a tool called Lean, which works like a strict robot referee for math: it checks every single logical step for errors, rather than trusting a human to read it over. To get there, OpenAI says its model first cracked a simpler version of the problem using about 1,000 AI “helpers” working together over 50 hours, then scaled up to roughly 10,000 helpers to go after the full problem.
Martin Bridson, president of the Clay Mathematics Institute (the group that created the $1 million prize), called it an exciting day for mathematics.
Step 5: It Wasn’t a Solo Race
Here’s where it gets messy. OpenAI wasn’t the only team chasing this. Just a day earlier, two human mathematicians, Levent Alpöge at Harvard and Tristan Buckmaster at NYU, released their own proof of a blow up, though for a simpler cousin of the equations that ignores friction entirely. They built their proof using AI tools too, including Anthropic’s Claude models. They said a solution to the full problem was coming soon.
On that same day, a separate team at Caltech, led by Anima Anandkumar, released yet another version of the simpler result, using a different kind of AI built specifically to understand physics.
Buckmaster publicly said he believed OpenAI learned about his and Alpöge’s approach shortly before releasing its own result, and copied the method. OpenAI has denied this, saying their AI reached the simpler result independently, through a different path.
Well known mathematician Terence Tao called Alpöge and Buckmaster’s work “a remarkable achievement” in its own right, a reminder that, headlines aside, real humans were racing toward this finish line too, using AI as a tool rather than letting it work alone.
Step 6: What Does This Actually Mean for You?
Don’t worry. Airplanes aren’t going to fall out of the sky, and your weather app isn’t suddenly wrong. The blow up scenarios (if confirmed) need extremely specific, unusual starting conditions that don’t resemble everyday wind or water. The Navier-Stokes equations will keep working great for virtually everything engineers use them for.
What this result would give us is a deeper map of exactly where a hugely useful equation and the real world can, in theory, disagree. That’s valuable insight for anyone studying turbulence.
It’s also a landmark for AI in mathematics. But it’s worth staying a little skeptical for now. This is a brand new, extraordinary claim, released amid a real dispute over credit, by systems nobody fully understands yet. Formal tools like Lean help (a computer double checking every logical step is a real safeguard), but the wider math community will want to pick this proof apart before anyone hands out a $1 million check.
The Bottom Line
A 200 year old equation that quietly runs the modern world just became the center of a very modern story. AI systems, not lone geniuses at a chalkboard, are racing each other to answer one of math’s biggest open questions. Whether it’s OpenAI, Alpöge and Buckmaster, or some mix of everyone who ultimately gets the credit, one thing is already clear: the fastest path to cracking math’s hardest problems may no longer run through a single human mind