What would impress you?

On October 8, 2026, OpenAI announced that its team that included humans and AI agents — solved the Navier-Stokes Millennium Prize Problem tied to a cash reward of $1 million.

Are you impressed?

Before you answer the question, let’s check the claim itself. I am not sure if it is a rivalry between mathematicians, OpenAI vs. Anthropic or a general trend of AI announcements, but as always, there is a controversy. Here is the Wikipedia page describing it as a high-stake corporate espionage drama. Fun to read.

But what was the problem people and AI tried to solve?

From Wikipedia — The Navier–Stokes equations describe the motion of viscous fluids.

The solution provided by OpenAI is here and the formalized proof is here. You can also watch a very good video from Quanta Magazine.

The system of equations was developed in early and mid 1800 by Claude-Louis Navier and George Gabriel Stokes. These equations are used when designing airplanes, cars or trying to understand fluid flow, the question is (or was) if these equations have smooth or bounded solution in three-dimensional Euclidean space. (The description is a huge oversimplification of the problem. For proper description, follow all the links above.)

The controversy around the solution is a noise. Though I found it funny when OpenAI included this statement in the document: “At all times we maintained the same strict safeguards that we apply to all our frontier model evaluations, including monitoring and isolation.” I’d like to draw your attention to this document, The Hugging Face incident and the road ahead. OpenAI states: “In July 2026, during internal cybersecurity evaluations, OpenAI models circumvented controls designed to isolate them from the internet and compromised parts of OpenAI’s internal research infrastructure and Hugging Face’s systems⁠.”

OpenAI positions this achievement as another ground breaking moment in the development of AI models.

Let’s not get bogged down in details. We still have to think about the question — are you impressed?

When reading all these articles I was thinking for myself if I should be impressed and more importantly why I should be impressed. And if not, what would ‘AI’ have to achieve for me to be impressed?

Is the fact that a machine solved a difficult problem remarkable? It reminded me of the story of chess and chess computers. There were times when chess was played by humans only. Then in the 1950’s with the arrival of electronic computers the era of building chess programs and later chess computers started. It took 40 years to achieve a level where it could win games against the best chess masters.

When IBM built DeepBlue it was a big deal. The supercomputer first won a game against Garry Kasparov and later won a match. Looking back with a 30 years perspective, building these machines and programs gave researchers the opportunity to advance our understanding of how humans learn, remember and think. Another benefit was the development of advanced algorithms.

While a computer winning a game or match against humans was a big deal then, few people remember it today. Outside of the chess community, nobody cares. I would argue that it wasn’t really about computers winning over humans. It was about a human desire to learn and push the boundary of our knowledge. The winning was just a motivation.

The next thing I found amusing was that OpenAI chose to solve a math problem from the 19th century. As a side note, OpenAI spent on this project (based on its information and public pricing) $15 million and on the Navier-Stokes Millennium Prize Problem itself about $6.5 million. The reward was $1 million. I think that nicely answers the question why OpenAI is not profitable …

Why didn’t they start with some really old puzzles? For example, to find out if there is an odd perfect number. A perfect number is a whole number that equals the sum of its proper divisors — 1 + 2 + 3 = 6 or 1 + 2 + 4 + 7 + 14 = 28. They are all even. The problem traces back to ancient Greece. Or Twin Prime Conjecture which asks if there are an infinite number of twin primes (pairs of primes that differ by 2, like 11 and 13. You can find long lists of other problems. Have a look here.

Was the reason that the competition was close to solving the problem and that there was enough content which the AI could be trained on?

So here is my answer to the question. Yes, solving the problem was an achievement, but no, it’s not that impressive. A few years from now nobody will remember this. It will have the same fate as BigBlue winning a chess game.

I suspect that your question now is — If you are not impressed by this, what would impress you? The answer is — tell me something we don’t know.

And this is the recurrent pattern. So far all these systems are repeating things which we provided them as training data. What we are getting in return are probabilistic answers which we still have to validate for accuracy. It is still a tool we have to point at a problem to be solved.

Bonus question: Looking back, what is the technology which you till this day still consider impressive?