Excellent People Working There
The easiest way to get an afternoon out of me is to be wrong about something I know.
Ask me to help build the sum of human knowledge and I will need to check my availability. Put the wrong result beside an old boxing match and suddenly I have no other obligations. I have become a man of public spirit. I have also become a man who intends to make a stranger admit he is wrong.
There must be some way to use this bastard.
Richard Cooke has been writing about the people who built Wikipedia. In an October 13 WIRED excerpt from his new history of the site, he describes an encyclopedia losing readers while its knowledge travels through AI answers and search summaries. The material can still reach you. The people maintaining it may not.
One of his sources, longtime Wikimedian Christophe Henner, gives me a recruitment method I recognize. In a personal essay first published in January, Henner describes someone reading an article, noticing an error or a gap, clicking “edit,” and becoming a contributor.
No stirring pledge required. Something was wrong. A person had five minutes. Five minutes is an estimate made before the person discovers what he has got himself into.
Here is an invented example, not a Wikipedia edit I have made or a chatbot I have tested. Suppose I look up an old fighter and find that he lost a particular bout on points. I remember his corner throwing in the towel. I have been carrying this useless piece of information for years, and at last somebody has supplied a use for it.
I want to change three words.
But my memory isn’t a source. The other fellow cannot inspect the little film running behind my forehead. Perhaps I remember another fight. Perhaps I have improved this one in the telling. So I need the report. I have to go looking through something old, find the right names and date, read farther than the bit that appears to agree with me.
This is work. It is also something I would cheerfully do to win an argument which nobody had asked me to enter.
The correction would belong on the page. Not in my private collection of victories. Someone looking up the fighter tomorrow could get the better account without knowing that I arrived in a disagreeable mood. My character needn’t improve before the sentence does.
Now put the same wrong result in an AI answer. I can object there too. The machine might apologize, give the correct result, and thank me for catching it. A pleasant arrangement. I have won immediately. Nobody has asked where I found the newspaper.
That doesn’t mean every system behaves this way, or that feedback can never reach the people maintaining it. It means the corrected conversation need not correct the shared article. The machine can satisfy me without taking me to the place where another reader could inherit the work.
I could click through a source link. I could hunt down the article myself. Nothing prevents me. But I began by wanting the result, and now I have it, with an apology thrown in. My afternoon has been returned to me before I developed a reason to spend it.
There is an even less scandalous version: the AI gets the fight right the first time. Fine. No bad information. No fake reference. I learn what happened and leave. As an answer, it succeeds. As an accidental invitation to join the people keeping the record, it has done nothing.
Henner’s worry is about the arrangement over time: fewer visitors can mean fewer people arriving at the shared pages who might become editors. In his account, the knowledge can spread more widely while the place producing it loses the encounters that help keep it going.
He doesn’t blame the entire decline on AI. He points to changing habits of reading and searching, to younger people consuming fragments, and to an encyclopedia that failed to reach many of the people newly coming online. His phrase is “AI is an accelerant, not the fire.” The trouble didn’t begin when a chatbot learned to answer.
Nor do I imagine the old arrangement was paradise for the newcomer. Henner also describes good-faith contributions being rejected, especially from people outside the familiar crowd. An edit button is not a guarantee that the people behind it will welcome you. I could arrive with my newspaper and find that I had joined an argument much older than mine.
But let this particular imaginary correction work. I find the report. It records a stoppage, not a decision on points. I put the source beside the changed result. Another editor checks it and leaves the correction standing. No conversion. No medal. A few words on a page about a man neither of us is going to meet.
I would have preferred an apology. At least a little recognition of the quality of my memory. Instead, tomorrow’s reader gets the result, and most of what it cost me is absent from the sentence. Nobody has to listen to the story of how I found the report. They don’t even have to know there was an argument available.
I read the entry again. The result is right. The source is there. It is a good deal less flattering than an immediate thank-you from a machine, and I am thoroughly pleased with it.
Very reliable encyclopedia.
Excellent people working there.
Primary essay: Wikipedia at 25: A Wake-Up Call (Christophe Henner; first published January 9, 2026; republished January 15).
Source: Wikipedians Believed AI Would Come and Go. They Were Wrong