Somewhere this week, a mathematician woke up, made coffee, and read the news that the problem they'd given their life to had been finished over a long weekend by a machine.
I keep thinking about that morning. Not the announcement — the morning. The specific, quiet vertigo of it. You've spent fifteen, maybe twenty years inside a single question. Not working on it the way you work on a task — living inside it, the way you live inside a language or a marriage. It's shaped how you think. It's the thing you were reaching toward on the good days and the thing that humbled you on the bad ones. And then one Tuesday, you open your laptop and it's just… done. By something that has never been alive, never lost sleep, never felt the particular ache of being close.
I don't want to write about the proof. Almost everyone else is doing that, and honestly, the proof isn't the interesting part. The interesting part is the person with the cold coffee. Because what they felt this week is something a lot of us are going to feel — quieter, smaller, but the same shape — and I think it's worth sitting with honestly instead of rushing past it.
What happened, quickly
Here's the doorway, and then we'll leave the math behind.
This week an unreleased AI model — running something like ten thousand agents in parallel — produced a 165-page proof of the Navier-Stokes problem, one of the seven great unsolved problems in mathematics. It's a question about whether fluids, described by certain equations, can "blow up" in finite time. It had stood for roughly ninety years. The machine did it in about 88 hours. The proof was checked by a system that mechanically verifies each logical step, which is a big part of why people are taking it seriously.
It's a few days old. Mathematicians are still scrutinizing it, and there's already a bitter dispute over credit. So hold it loosely as a result. But as a moment — as a thing that happened to real people — it's already real enough to talk about.
Because ninety years of human reaching just met 88 hours of machine, and some of the humans who were reaching are having a very hard week.
Sit with them for a minute
What do you actually feel, when the thing you gave your life to is finished by something that doesn't know what a life is?
I don't think it's one feeling. I think it's several, and they don't sit comfortably together.
There's grief — but not for the problem. For the relationship with the problem. You don't spend two decades with a question without it becoming part of who you are. It's not your job; it's your companion. And when it's suddenly answered — by something else, in a weekend — there's a loss that has nothing to do with prize money or citations. It's the loss of the thing you were going to spend the rest of your life reaching toward. The reaching is over now. Someone turned on the lights.
There's the credit wound, which is sharper and more human than it sounds. One of the mathematicians was mid-way through his own related work when the announcement landed, and he's accused the lab of muscling in. Imagine that — not just being outpaced, but being outpaced on the exact thing you were holding in your hands right at that moment. That's not abstract. That's someone watching a door they were walking through get walked through by someone else, faster, while they were still in the frame.
And there's the worst combination of all: awe and resentment in the same breath. Because if you love this stuff — and these people love this stuff — some part of you looks at what the machine did and thinks my god, that's beautiful. And in the very next beat: and it cost me everything, and it cost that thing nothing. You can't even cleanly hate it. That's the cruelty of it. The thing that hollowed you out is also, genuinely, magnificent.
Underneath all of it is a question I can't stop turning over: was the point the answer, or the chase? Because if a machine can just hand you the answer, then what were the twenty years for?
Now look in the mirror
Here's the thing, and you probably already feel it: this isn't a story about mathematicians.
You're reading this and something in your chest is doing something, and it isn't about fluid dynamics. You don't care about the Navier-Stokes equations. You care because you can see it coming for your thing. The thing you're good at. The thing you spent years getting good at — the late nights, the concepts that finally clicked at 2 a.m., the slow accumulation of a skill that made you feel like you were someone who could do this.
The mathematician is just the canary. They went into the mine first, and they felt the full dose of a gas the rest of us are only starting to smell. Every skilled person alive right now is getting a low, background version of the exact feeling that hit those mathematicians at full strength this week: the quiet sense of the ground shifting under the thing you're good at.
We've been talking about this in the language of jobs and productivity and disruption, because that language is safe. But that's not what the feeling actually is. Let me try to say the real thing.
What the grief is actually about
The fear isn't really "I'll lose my job." That's the acceptable version, the one you can say at a standup.
The real one is quieter and goes deeper: the fear that the hard-won thing — the skill you suffered for, the years of difficulty you're secretly proud of — might stop mattering.
Think about what mastery costs. Nobody gets good at anything worth doing without a stretch of being bad at it, frustrated by it, humbled by it, and slowly — painfully — becoming someone who can do the thing. That process changes you. The difficulty is why the skill feels like part of your identity: you earned it, and it cost you something, and the cost is what made it yours.
And the fear underneath this whole moment is that all of that — the cost, the earning, the identity built on top of it — could become a museum piece. A thing that used to be hard and impressive, that a machine now does casually, the way a calculator made "good at long division" into a sentence that means nothing. Not that you'll be unemployed. That you'll be unremarkable at the thing that made you feel remarkable.
That's a real grief. And I don't think it helps anyone to wave it away with "just learn to use the tools." That's true and it's also not the point, and everyone can tell when you skip the feeling to get to the advice. So I'm not going to skip it. It's hard. It's allowed to be hard.
What the machine didn't get
But sit here long enough and something else comes into focus, and it's the thing I actually want to leave you with.
Most of those mathematicians never expected to be the one who solved Navier-Stokes. They're not naive — they knew the odds. The problem had eaten ninety years of brilliant people. Statistically, they were almost certainly going to spend their whole careers reaching for something they'd never quite reach.
And they did it anyway.
Which means the thing that kept them going was never going to be the one who finishes it. It couldn't have been — the math didn't support that hope. What kept them going was the years inside a beautiful, impossible problem. The reaching itself. The way a hard question organizes a life, gives your attention somewhere worthy to go, makes you into a particular kind of person — patient, humble, awake to something larger than you.
The machine got the answer. It did not get the twenty years of being someone who was reaching for it.
That's not a consolation prize, and I don't mean it as one. It's the actual distinction. The output was never the thing that made a life. The output is a page. The life was the reaching — the caring, the difficulty, the being-changed-by-the-work. A machine can take the deliverable. It genuinely cannot take the fact that the work made you into someone, because it was never trying to become anyone. It doesn't know there's anyone to become.
That doesn't make the grief fake. The grief is real, and it should be — you're mourning something that mattered. I just don't think what you're mourning is the part that got taken. The answer got taken. The becoming was always yours.
So
I don't have a tidy bow for this, and I'd be lying if I offered one. A machine finished a lifetime of work in days, and a version of that is coming for most of us, and it's going to feel strange and unfair and a little bit like grief when it does.
The mathematicians felt it first and loudest this week. I think the honest response isn't to reassure ourselves that we're irreplaceable, and it isn't to spiral into "we're all obsolete." It's to do the human thing the machine can't: to feel it, out loud, together, and to remember that the reaching was never wasted just because something else reached the end faster.
If you loved the work, the work already gave you the thing that mattered. It made you into someone. Nothing finishes that in 88 hours.
Have you felt it yet — the moment AI did the thing you were quietly proud of, and you didn't know how to feel about it? I don't think there's a right answer, and I'm not sure the answer is even the point. But I think we should say it out loud, instead of all pretending we're fine. I'll go first in the comments. Your turn, if you want it.
Top comments (34)
I want to add the morning nobody writes about. For every mathematician who watched a machine finish his life's work, there's a whole crowd on the other side of town waking up to the opposite morning — people who spent their lives outside every room like that. No degrees, no training, jobs that give nothing back. The same machine that emptied his morning is the only reason theirs have anything in them at all. I'm one of them, and this isn't me waving off the grief — it's real, and your line about becoming unremarkable at the thing that made you feel remarkable is the truest sentence in the piece. The door just swings both ways, and both mornings are true at once: the thing that creates things took the reaching from someone who had it, and handed a reaching to those who were never going to get one. We made cars to cross distances and electricity to see in the dark, and now we've made the thing that makes things — and this is only the beginning, for good and for bad, both. Nobody can tell the mathematician his loss isn't a loss. But somebody should tell him what it bought: people who were never going to be let in, building anyway. The becoming didn't end. It moved, and it multiplied. I'm willing to embrace that — not because it's safe, but because it's here, and I'd rather stand with the people it's making than let it pass by.
Tools change, purpose remains ✌️
This is the half I couldn't have written — thank you. Both mornings are true at once, and "the becoming didn't end, it moved and multiplied" is truer than anything in the piece.
Appreciate that. And since the question that got deleted was aimed at both of us — how do you recover your sense of existence when the machine takes your thing — I'll leave my half here anyway, because more people are wondering it than asking it. The meaning of existence is exactly what you put into it. Exist to serve people and their idea of what the world should look like, and your existence rides on the whims of others — one verdict can repossess it. Exist to follow your path and let the work decide the result, and you exist for yourself — there's nothing for anyone to repossess. And here's why that choice matters more now than ever: the machine is coming for what you do either way. Whether you use it or not, it exists, and like every invention since the beginning of time it will take jobs and self-worth — it took things from me before I ever used it. What it can't decide is whether it also gets what you are. That part depends entirely on where you kept your existence. Store it in the title, the salary, the crowd's judgment of your output — the machine takes the output and you go with it. Store it in the path, the work, the thing you can hold in your hands — it can take your job and still walk away hungry. I keep an actual research file of what happens to people who stored their whole selves in the machine's confident voice: careers, and worse. So that's my answer: place the value of your existence on something you can hold in your hands, not something people judge you from. That's the one thing it can't finish in 88 hours.
You've sharpened my point better than I made it: keep your worth in something you can hold in your hands, not something others judge you from — that's the exact line between what the machine can repossess and what it can't. "It can take your job and still walk away hungry" is going on my wall.
No, we actually first made trains for that; then "we" made cars for convenience, and now that we've grown reliant on them, now that some societies have structured themselves around cars so thoroughly, that not having one makes you a second class citizen, car companies are slowly, inch by inch, turning car ownership into a subscription.
This is my main problem with AI: People say it "democratizes" open source, but in reality, it turns participation into a service. And whenever the bubble bursts and AI gets expensive, the first to lose access will be the ones who needed it the most.
The vision is great; AI opens the door to everyone now, and we can all be mathematicians. I have nothing against this idea. But the reality is, the door hasn't opened, it's been turned into a subscription, and even after getting in, you need to keep paying if you don't want to be kicked out.
I agree with your point but the car line was one throwaway example in my list, not my point. my point was the two mornings.
and even on the car thing. Cars aren't just "trains for convenience" — trains only go where somebody already laid track. cars go where no track will ever run. That's not convenience bolted onto a train, that's a whole different capability.
But you're dead right about the part that matters: access is getting turned into a subscription, and when the bubble pops the first ones cut off are the people who needed it most. that's real and it's ugly. where I land different is I don't think that's the only ending — it's the default one, and the default only wins if nobody builds the other door.
No, for most people, cars can also only go where someone has already built a road. Setting aside some few cases, convenience really is the main factor. But that was really just a comment on the side, it doesn't really affect your point or mine.
More importantly, I just took up your car example because it was also a good example for what I wanted to say, so it just felt natural to keep going with cars.
Oh absolutely, I'm 100% with you there. But right now there's a lot of capital backing this outcome, while not a lot of people seem interested in averting it. We've seen in plenty of other products that people often just allow companies to turn anything into a subscription, so there's no reason to assume most people won't also just watch as skills like mathematics and programming are now also turned into a subscription.
It is, as you say, the default, so it's what will happen unless enough people make an active effort to avoid it. And right now, I don't really see that happening anywhere.
I agree with that too — but I think you're taking the effect as the whole truth without looking at the cause. The inventor of the automobile didn't create it for convenience; his reason was something different. The convenience came from the effect of it existing, not from why he built it. Same with AI: the money and the power are the effect. I'm not sure if that was the reason it was created initially, so I can only speculate.
It was nice to discuss it with you. This is what I like to talk about — the heavy stuff, the inner workings of people, not just harnesses and A.I. So anytime you want to open a thread, I'm all for it. Take care.
The mathematician's version is dramatic because it happens once, in public, with a name attached. Most of us won't get that clean a moment. The actual threat is quieter: the AI doesn't take your one big problem, it takes the thousand small reps that built the skill in the first place. You don't notice the loss because there's no single Tuesday to point to. Just one day you realize you haven't done the thing that used to make you good at the thing.
The quiet version is worse precisely because there's no Tuesday to grieve — you don't lose the skill, you stop building it, and notice only once it's already gone.
There are two things. Firstly, this is not everyday AI. It's an ensemble of 10,000 AI models that are ahead of GPT 6 Astra, running for about 88 hours to arrive at a solution. This group was one of many groups attempting it. They first solved an easier problem and then pointed their guns at this problem, taking non-obvious insights from it.
Secondly, you're complaining about why the universe supports greater forms of intelligence than you. This has always been the case. When calculators were invented, mathematicians were supposed to lose their jobs. When autopilots were invented, people suspected pilots would lose their jobs. Ironically, there are more jobs than ever in both these fields.
Being able to solve difficult problems and uncovering non-obvious insights with AI is exactly what would propel scientific discovery forward. Mathematicians and physicists would be busier than ever cracking deeper problems that were previously unchartered territory.
Fair on the scale — this was a 10,000-model ensemble, not everyday AI, and the calculator/autopilot precedent is real. But the piece wasn't complaining the universe allows greater intelligence; it was sitting with the grief of the people mid-reach. Both can be true: more problems open and a hard week for whoever was holding this one.
Maybe Jevons paradox applies to workers as well sometimes? Making workers more effective means companies will want to employ even more of them, because each individual one is now more effective, so profit is in scaling up rather than down...
I think there are three forces here.
First, the efficiency gain: AI reduces the number of humans needed per unit of work.
Second, the Jevons effect: lower costs make previously uneconomical startups, products, and projects viable. So average company size may shrink while the number of companies increases.
Third, saturation: demand isn’t infinite. We already have decent solutions to many everyday problems, and attention, money, distribution, etc. remain scarce.
So the real question is whether the increase in economically viable work outpaces the reduction in human labor needed to do each unit of work.
Yea, that does seem to be the interesting question here, and I'd say pretty confidently that nobody out there can really predict for sure which way it will go. Anyone claiming more than maybe 70% to 80% certainty would just seem suspicious to me.
There's also an added dimension here: AI work might also create new demand, be it by producing bad code that needs to be cleaned up, accelerating the framework hamster wheel, or whatever.
In the end, only time will tell.
The real thing is that most of those mathematicians knew eachother, or atleast of eachother... They expected them, or 1 of their peers to finish it, so they can rest assured that it was solved by someone they respected... In walks a machine that wiped the floor with their entire field... Not a great way to start off the week? Your life's work, compressed into 88 hours and all you can do is see if they made a mistake... But they didnt... 88 hours and you take longer to verify a subset of it than it took to do it. A stark reminder that AI will outpace us all in time.
You've named the quietest cut in it: spending longer to verify a subset than the machine took to do the whole thing, and finding no mistake to soften it. And if 88-hours-and-flawless is the floor, not the ceiling — shouldn't we be at least a little worried about what's left that's actually ours?
the credit wound part hits hardest for builders too. losing isn't the scary part, losing while mid-way through the exact same thing is. seen founders ship two weeks before a near-identical competitor launch and it stung worse than a whole bad quarter ever did.
The near-miss stings more than the loss — you can't tell yourself you weren't close.
A fascinating and at the same time melancholic thought. Mathematics (and deep scientific research in general) has long been our ultimate bastion of human meaning—a place where you make yourself immortal through years of lonely struggle. When a machine now short-circuits this path in record time, the value shifts from the solution to the search for meaning. We will sooner or later have to accept that the process of thinking has an existential value for us in itself, even if the end product can be efficiently automated. The real crisis is not an economic one, but a crisis of human purpose.
"A crisis of purpose, not economics" is the truest framing in this thread. If the process of thinking has existential value even when the product is automated, then meaning was never in the output — and that's the only ground the machine can't take.
What gets repriced in that morning isn't the proof - it's where judgment sits. For twenty years the scarce thing was producing the argument; now the scarce thing is knowing whether the argument is right, and knowing which question was worth twenty years. Both of those stay with the mathematician, but they're different work than the work she trained for, and the training didn't exactly prepare her for the morning after. The optimistic read: the person who lived inside the problem is now the best verification engine the machine's output will ever face, and verification is suddenly the bottleneck everywhere - in math, in code, in everything else the machines finish in days. The vertigo is real and deserves honesty. But the skill that causes the vertigo is also the one the new world is shortest on.
"What gets repriced isn't the proof — it's where judgment sits" reframes the whole grief precisely: producing the argument was the scarce thing, now knowing if it's right is. And the honest optimism is real — the person who lived inside the problem is the best verifier its answer will ever face — but you named the catch too: the training didn't prepare her for the morning after.
This line stopped me in my tracks: "The machine got the answer. It did not get the twenty years of being someone who was reaching for it."
As developers, many of us are quietly going through a micro-version of this exact morning. You spend a decade developing deep intuition for concurrency, memory leaks, or tricky distributed edge cases—the things that required genuine mental scars to earn. Then, a model synthesizes a working solution in 4 seconds.
The initial reaction is that hollow sting of: "Was all that friction pointless?"
But reading this reminded me of something critical: friction is where judgment is formed. The model produces syntax and answers, but it doesn't possess the scars, the context, or the taste that tells you why an architecture feels right for actual human beings.
The machine can hand us the finish line, but it can’t experience the craft. Thank you for putting words to the quiet grief so many builders are feeling right now.
"Friction is where judgment is formed" — that's the whole essay compressed better than I wrote it. The model skips the friction, which is exactly why it gets the answer but not the taste the scars taught you. Thank you for this.
I really appreciate this reflection. It shifts the focus from competing with speed to leaning into the nuance and craftsmanship that only comes from years of dedicated practice and human perspective.
Thank you — and that's exactly the shift I was reaching for: the moment you stop racing the machine on speed, you notice the thing it was never in the race for was the craft all along.
In the simplest form, this is the story of the hero losing his superpowers or magic sword or whatever his story uses to externalize his specialness. It's an old tale and probably just as old a fear.
As humans we like the idea that status is inalienable; your house can burn down, your money can get stolen, but if you had it in you before to get those things, then you sill do and you can get them back just the same. Having the qualities that give us this sense of safety be taken away is, understandably terrifying, and even more so when those qualities are earned, at least partly, with hard work.
Here's my take: When I come home after work today, I will probably have some tea, and before that, I'll have the choice of brewing vessel. If I just want my tea then move on with the day, I'll likely pick a relatively cheap, mass produced glass gaiwan. It's simple, perfectly round in a way only a machine could produce, and just overall unremarakable. No human could create such a perfectly consistent product for such cheap a price. If, however, I want to spend take some more time and really enjoy the process, I'll take one of my more expensive, hand crafted tea pots off the shelf, get some nicer tea, and simply appreciate that I'm working with something made by a real human, by hand, without a single identical one on the whole planet.
As AI takes over more industries, people will just have to develop that kind of appreciation for more aspects of life. Where human-made used to be the only way, in the future, it will simply be the artisanal option. A clanker will build you a website when all you want is a usable web view over some data, and a real person might build you a personal site if you want something a bit more special, with a human touch, that one can appreciate for its own sake.
Oh, and we need to abolish capitalism. Like, preferably yesterday. Otherwise, forget all of the above, the future is slavery under AI feudalism and those who provide no value don't get to eat.
Some comments may only be visible to logged-in visitors. Sign in to view all comments.