The Commons Is Going Quiet
Almost everything I know about building software I learned from people I have never met. Not from a course and not from a manual, but from a stranger somewhere who took twenty minutes to write out an answer to a question they had no obligation to answer, for a person they would never hear from again, usually at an hour when they should have been asleep. Multiply that by a couple of decades and you have my education. You probably have yours too.
I want to be precise about what that was, because we have started to lose it and I do not think we are being honest about why.
It was not charity. It was an economy. Nobody was paid in money, which is exactly why the whole thing looked free and therefore looked permanent. The payment was subtler and, for a certain kind of person, more powerful: you were read. Your name sat under the thing you wrote. Someone replied. Someone disagreed and you got sharper. A hiring manager found your answer three years later and it changed your life. The commons of the open web ran on attribution, reputation, and the ordinary human pleasure of being useful in public.
We have spent the last few years building extraordinary machines on top of that commons. And in the process we have quietly stopped paying into it.
Not theft. Evaporation.
The argument I keep hearing about this is a copyright argument, and I think it is the least interesting version of the problem. Set the legal question aside entirely — assume every training question gets settled tomorrow in a way everyone can live with. The structural problem does not move.
Here is the structural problem. When you ask a machine and the machine answers, the answer arrives smooth, synthesized, and sourceless. It is often excellent. It saves you an hour. But the person whose careful post made that answer possible receives the one thing they were never willing to work for free in exchange for: nothing. No visit. No reply. No name. Not because anyone took something from them, but because the room they were writing into has emptied out.
That is not theft. It is evaporation. And evaporation is harder to argue about, because there is no villain in it and no moment where it happens.
Ask the question the way an operator would: what is the return on writing a thorough public answer in 2026? A few years ago the honest answer was a real one. Today, for most people, the honest answer is that your words become invisible infrastructure inside someone else's product. Some people will keep writing anyway, because they always would have. Most will not. Most people are not saints; they are reasonable, and reasonable people respond to incentives.
The dataset is the smaller loss
The version of this worry that gets the most attention is about model quality — systems trained increasingly on the output of systems, a photocopy of a photocopy, the corpus slowly going stale. That is a real concern and the labs are not unaware of it. It is also, I think, the second most important thing happening.
The first is that we are dismantling the place where people became competent in front of each other.
The public forum was never only a database. It was an apprenticeship, conducted in the open, at scale, for free. You asked a naive question and someone senior corrected you, and forty other people who never asked read the correction and learned it too. You watched experts argue and absorbed what a good argument even looks like in your field. You learned taste, which is the thing no documentation contains. Then, some years in, you found yourself writing the answer instead of asking, and that turn is how a profession renews itself.
A private conversation with a model gives you the answer. It does not give you the room. And the room is where the next generation of people who could have written the answer used to be standing.
What I am not saying
I am not nostalgic, and I am not making the case that we should have kept things the way they were. I use these tools every day, in this business, on real work. They are the most useful thinking partners I have ever had, and the number of hours they have given back to me is not small. Anyone who tells you to refuse them is asking you to work with one hand tied for the sake of an aesthetic.
What I am saying is that we are drawing from a well faster than it refills, and calling the water free because the bucket is cheap. That is not a moral complaint. It is an accounting one. Every system that consumes a shared resource without a replenishment path eventually meets the same wall, and the wall is usually a surprise only to the people who were not counting.
Three things that would actually help
The interesting question is not whether to use the tools. It is how a synthesizing layer can sit on top of a commons without starving it. A few things seem obviously right to me, and none of them require anyone to be heroic.
Send traffic back. Not a footnote nobody clicks — a real path to the source, presented as worth taking. If the source gets nothing from being helpful, it will stop being helpful, and then the product built on top of it degrades. This is self-interest, not generosity.
Cite legibly. Attribution was the currency. Paying it costs almost nothing and it is the difference between a contributor feeling drawn from and feeling drawn on. Those are not the same word.
Design products that end somewhere else. This is the hard one, because every incentive in software points the other direction. A tool that answers your question and then hands you to a person, a community, a book, a place where the learning continues, is a tool that gave up some session time on purpose. I think that is the whole design problem of this decade, and I think most companies will get it wrong for entirely rational reasons.
The commons was never a natural resource. It was a habit — millions of people choosing, over and over, to put something useful where a stranger could find it, because putting it there came back to them in ways that were real even if they never showed up on a balance sheet. Habits are more fragile than resources. You do not deplete a habit. You just stop giving people a reason, and one day you notice how quiet it has gotten.