Guardians of the Sum
On the free encyclopedia, the people who keep it standing, and the strange thing that happens to power when it has nowhere else to go.
Somewhere right now, a person who has never edited Wikipedia before is fixing a sentence. Maybe they noticed their hometown’s article gave the wrong founding year, or that the page on a beloved obscure novelist misspelled her married name, or that a citation pointed to a dead link and the claim it once supported had quietly gone feral. They are not a vandal. They are not pushing a cause. They are doing the small civic thing the whole project was built to invite — the thing printed right there on the front door, in letters every newcomer reads as a promise: the free encyclopedia that anyone can edit. They hit save. They feel, for a moment, like a participant in something enormous.
And then, often within minutes, sometimes within seconds, a stranger they will never meet undoes the entire thing.
Not with a note. Not with welcome, and thanks, but here’s the issue. With a terse line in an edit summary, a string of capital letters and colons that means nothing to the uninitiated — rv unsourced, see WP:V, WP:RS — and the sentence is gone, the founding year wrong again, the promise quietly amended in the fine print to read: the free encyclopedia that anyone can edit, provided they already know the rules they were never told. The newcomer stares at the screen. They were trying to help. They came in good faith. And the very first thing the encyclopedia did, through one of its keepers, was treat them like a threat.
This essay is about those keepers. It is not an attack on them — or it is not only that, and I want to be honest about the difference, because the honesty is the whole point. I love Wikipedia. I mean that without irony and without the defensive cough people put before a complaint. It is one of the few genuinely good things the internet has produced, a cathedral built by strangers, free of charge, out of nothing but the conviction that knowledge should belong to everyone. It should not exist. It exists anyway. And it exists because of an army of unpaid volunteers who spend their evenings reverting graffiti, chasing down sources, and standing between the sum of human knowledge and the endless tide of people who would scrawl obscenities across it for a laugh. Strip those people out and the cathedral becomes a latrine wall inside a week. Their necessity is not in question. It is, if anything, the thing that makes the rest of this so hard to say.
Because here is the trouble with necessary power: it is the kind that is hardest to criticize, and therefore the kind that least often gets criticized, and therefore the kind most likely to rot.
The autoimmune turn
Think of moderation as an immune system. That is, more or less, what it is. A healthy body needs one — needs cells whose entire job is to recognize what doesn’t belong and remove it, fast, without sentiment. A body without an immune system dies of the first thing that wanders in. Wikipedia’s moderators are its white blood cells: they identify the foreign matter — the spam, the libel, the slow-drip vandalism, the marketing department quietly editing its own CEO’s biography — and they neutralize it. The system works astonishingly well. The fact that you can look up almost anything and find it mostly correct, mostly current, mostly free of someone’s pet conspiracy, is a daily miracle of immune function performed by people you will never thank.
But immune systems have a failure mode, and it is not weakness. It is the opposite of weakness. The failure mode is when the system, tuned for years to attack, loses the ability to tell the threat from the body. It begins treating healthy tissue as an invader. It attacks the thing it was built to protect. We call that autoimmune disease, and the cruel signature of it is that the symptoms look like vigor — the system is working harder than ever, mounting response after response, except now the responses are the illness. The patient is being destroyed by their own defenses, and the defenses experience this as doing their job extremely well.
That is the diagnosis I want to make about Wikipedia’s culture of moderation, and I want to be precise that it is a cultural and structural diagnosis, not a moral one about individuals. The newcomer with the founding year is not a threat. The retiree adding a paragraph to her town’s history is not a threat. The grad student citing a real paper that happens to violate an obscure formatting convention is not a threat. They are healthy tissue. And too often, the immune system has been running hot for so long that it can no longer tell. It reverts first and asks nothing later. It has forgotten that it exists to protect the body of contributors, and has come to experience that body as the thing it must defend the encyclopedia from.
The rest of this essay is, in a sense, just an attempt to describe the symptoms of the autoimmune turn — and then, because I am not interested in a hit piece, to say plainly what the immune system would say in its own defense, which is more than you might expect, and which any serious critique has to survive.
The cudgel made of letters
Start with the language, because the language is where you first feel it.
Wikipedia runs on an internal dialect of shorthand — hundreds of policy and guideline pages, each with a memorable abbreviation. WP:V is verifiability. WP:RS, reliable sources. WP:NPOV, neutral point of view. WP:NOR, no original research. WP:N, notability. WP:BLP, the rules for articles about living people. There are hundreds more, a whole alphabet of governance, and in their proper use they are a genuine marvel: a shared constitution that lets thousands of strangers coordinate without ever speaking, a way to say here is the principle at stake in four characters instead of four paragraphs. The shorthand is not the problem. The shorthand is, in fact, one of the most impressive pieces of social technology the open web ever produced.
The problem is what the shorthand becomes in the wrong hands, which is a cudgel. Watch a contested talk page long enough and you will see it: a discussion that should be about whether a claim is true and well-sourced gets transmuted, almost magically, into a discussion about which policy acronyms can be deployed most fluently. A newcomer makes a substantive point — a good one, a sourced one — and the response does not engage the substance at all. The response is a wall of links. Per WP:UNDUE and WP:FRINGE, and see also WP:ONUS — the burden is on you. The newcomer, who does not yet speak the dialect, has two options. They can go read all of it, which takes hours and is exactly what the veteran is counting on, or they can give up, which is also exactly what the veteran is counting on. Either way the substance — is the claim true? — never gets discussed. It has been buried under procedure.
Wikipedians have a word for this. They call it wikilawyering, and the fact that the community had to invent the word tells you the behavior is real and recognized. Wikilawyering is the use of the letter of the rules to defeat their spirit — quoting policy to win rather than to clarify, treating the encyclopedia’s governance as a game of jurisdiction rather than a shared search for accuracy. Everybody on Wikipedia agrees wikilawyering is bad. Everybody can name someone else who does it. And it goes on constantly, because the structure rewards it: in a dispute, the person who knows the most policy pages and can cite them fastest very often wins, regardless of who is actually right about the thing under discussion. We have built a court where the verdict goes to whoever has memorized the most case law, and then we are surprised that the building fills up with lawyers.
What this does to discourse is subtle and corrosive. It establishes, without anyone ever quite saying it, that fluency in procedure outranks knowledge of the subject. The actual historian, the actual physician, the actual person who was there, arrives with expertise and is told, in effect, that their expertise is worth nothing until they can express it in the local liturgy. Sometimes that gatekeeping protects the encyclopedia from cranks — and we’ll come back to that, because it genuinely does. But just as often it functions as a humiliation ritual, a way of informing the outsider that they are an outsider, that the knowledge they brought is the cheap part, and that the real currency here is not what you know but whether you know the secret handshake. The pursuit of rigor mutates, slowly, into a defense of the priesthood.
My article
There is a guideline on Wikipedia called WP:OWN, and it is one of the most quietly damning documents the project has ever produced — damning not because of what it says but because of why it had to be written. The guideline exists to remind editors that no one owns an article. That you cannot stake a claim to a page, hover over it, and treat every change made by anyone else as a trespass on your territory. That you do not, in fact, get to plant a flag in the entry on some battleship or some obscure mathematical lemma or some mid-tier rock band and rule it as a private fiefdom.
The guideline exists because this is precisely what a certain kind of veteran editor does.
You learn to recognize the territory the way you learn to recognize a guard dog’s yard. Certain pages have a keeper — someone whose watchlist pings the instant anyone touches it, who has appointed themselves the sole arbiter of what the article may say, and who reverts incoming changes with a speed and reflexiveness that has nothing to do with the merit of the change and everything to do with who made it. A stranger improves a sentence; it is gone in ninety seconds. The keeper did not weigh it. The keeper does not need to weigh it. The keeper has simply decided that the page is theirs, and the mere fact of an outside hand touching it is sufficient cause for reversal. They will cite a policy, of course — there is always a policy — but the policy is the costume, not the motive. The motive is ownership, and ownership is precisely the thing the encyclopedia is supposed to forbid.
This is the “mafia” feeling, if we are being honest about where that word comes from. Not organized crime — nobody is running numbers. It is the feeling of a territory with informal bosses, of turf you can wander into without knowing you’ve crossed a line, of a protection arrangement you never agreed to and cannot opt out of. You came to add a fact. You did not realize the fact lived in somebody’s neighborhood. And the resident has ways of making it clear that this is his block, that he was here first, and that you are a guest whose welcome is entirely at his discretion — on a platform whose entire stated identity is that there are no guests, only editors, all of them equal, all of them invited.
Victory by exhaustion
The most important thing to understand about how power actually operates on Wikipedia is that it is rarely a matter of a single decisive blow. It is a matter of time, and specifically of the brutal asymmetry of time between the veteran and the newcomer.
Consider the mechanics. Wikipedia has a rule called the three-revert rule — 3RR — which forbids any editor from undoing the same content more than three times in a day. It exists to stop edit wars, those grinding back-and-forth reversion duels where two people undo each other into infinity. On paper it is perfectly even-handed: it binds everyone equally. In practice, the veteran has learned to fight just under the line. Three reverts today, three tomorrow, the pace of attrition calibrated precisely to stay technically legal while making the page a misery to touch. The newcomer, who does not know the rule exists, hits revert a fourth time in genuine confusion about why their correct edit keeps vanishing — and that, the newcomer’s fourth revert, is the one that gets reported. The veteran, who has played this exact game a thousand times, has the procedural fluency to walk the newcomer straight into a violation and then file the complaint. It is not that the rules don’t apply to the powerful. It is that the powerful know the rules well enough to weaponize them, and the powerless do not know the rules well enough to survive them.
And underneath the rule-gaming is the deeper asymmetry, which is simply that the veteran has nothing else to do here and the newcomer has a life. The veteran has tooling the newcomer can’t see — one-click rollback, automated watchlists spanning thousands of pages, scripts that surface every change to their territory in real time. The veteran has patience, the specific infinite patience of someone for whom this is the main event, who will outlast you across days and weeks and months of talk-page argument because the argument is, for them, the point. The newcomer came to fix a founding year. They did not come to litigate the founding year for three weeks against an opponent who treats the litigation as a hobby. So they leave. Not because they were wrong. Because they were tired, and the other person was not, and the other person knew they would be.
This is what victory looks like on Wikipedia far more often than any dramatic ban or block: it looks like a good editor quietly closing the tab and never coming back, beaten not by a better argument but by a better-resourced one, ground down by an opponent with more time, more tooling, and more appetite for the grind. The encyclopedia does not register this as a loss. There is no log entry for contributor of genuine value, exhausted into departure. The page stabilizes. The keeper’s version stands. The immune system records a success. And somewhere a person who would have made the encyclopedia a little more complete decides, correctly, that it is not worth it.
The noticeboard as theater
When a conflict can’t be settled by attrition, it goes to the boards — the noticeboards where disputes are aired, the most notorious of which is the Administrators’ Noticeboard for Incidents, universally known by its initials and universally dreaded. In theory these are venues of justice: a neutral place to bring a problem and have experienced community members weigh in. Sometimes that is exactly what they are.
But a noticeboard is also a stage, and a certain kind of dispute on Wikipedia is less a trial than a performance with an audience. The veteran who brings a newcomer to the board does not arrive alone. They arrive with history — with allies they’ve edited alongside for years, with a reputation, with the immense and rarely-acknowledged advantage of being a known quantity in a room full of known quantities, facing a stranger nobody has any reason to extend the benefit of the doubt to. The discussion that follows is nominally about the facts of the dispute. It is functionally about belonging. The regulars recognize one of their own and recognize an outsider, and the gravity of the room pulls accordingly, and it pulls before a single piece of evidence has been weighed. The newcomer experiences this as a pile-on, because that is what it is, and they are told — they are always told — that they are being uncivil for noticing.
WP:CIVIL is, by the way, another real guideline, and it deserves a special place in this essay, because the civility policy is where the autoimmune turn shows its hand most clearly. Civility is enforced, on Wikipedia, with a striking selectivity. The veteran who reverts a newcomer twelve times with curt contempt and never once explains himself is rarely cited for incivility; he is, after all, just enforcing standards. The newcomer who finally, after the twelfth revert, types why do you keep undoing this, I gave you a source — that can get flagged as a failure of civility, an unbecoming display of frustration, evidence of an editor who “can’t collaborate.” The cold contempt of the powerful reads as professionalism. The hot frustration of the powerless reads as a conduct problem. Same temperature of feeling, opposite reading, and the difference is not the behavior. The difference is who is doing it to whom.
Here is the irony I cannot get past, the one that sits at the dead center of all of this. Wikipedia already has the rules to prevent everything I am describing. There is a guideline literally titled “Please do not bite the newcomers” — WP:BITE — instructing experienced editors to treat newcomers with patience and good will, to assume they are trying to help, to remember that everyone was new once. There is a foundational principle, WP:AGF, “assume good faith,” which asks every editor to begin from the presumption that the person across from them is acting in good faith until proven otherwise. There is the civility policy. There is the ownership guideline. The community looked, years ago, at exactly the failure modes this essay describes — the biting, the contempt, the territorial hoarding, the bad-faith assumption — and it wrote rules against all of them. They are on the books right now. Anyone can read them.
And they are the least enforced rules on the entire platform. The obscure formatting convention gets enforced in seconds. The citation-style preference gets enforced in seconds. “Do not bite the newcomers” gets enforced approximately never, because there is no one whose job it is to enforce it, and because the people who would be cited under it are, very often, the same people doing the citing. The rules that protect the encyclopedia from bad content are sharp and quick. The rules that protect people from bad treatment are decorative. That asymmetry is not an accident or an oversight. It is the autoimmune disease in its purest form: a system that has grown exquisitely good at attacking and forgotten entirely what it was supposed to protect.
What happens to power with nowhere to go
I have been careful, so far, to keep this structural — to talk about incentives and mechanics rather than to psychoanalyze the people involved, because the lazy version of this argument is a cheap shot and the cheap shot is both unkind and untrue. You have surely heard the lazy version. It says that Wikipedia’s worst moderators are sad people with no power in their real lives, compensating online for what they lack offline — that the rule-obsessed editor is a small man made temporarily large by an admin bit. It is a satisfying story to tell about someone who has just reverted your work for the fifth time. It is also, as a general theory of human beings, mostly garbage, and I won’t make it, because it lets the real culprit off the hook.
The real culprit is not the people. The real culprit is the structure, and the structure would produce these behaviors in almost anyone you fed into it, because of what kind of power it offers and what conditions it offers it under.
Consider what the power of a Wikipedia moderator actually is. It is real — the ability to revert, to block, to delete, to shape what millions of people will read when they look something up — and it is also, crucially, unaccountable in the specific sense that the people you exercise it over almost never can do anything about it. It is exercised under pseudonymity, so it carries no reputational cost in the life that matters most to you, the offline one. It is conferred and adjudicated within a closed status economy — a world of barnstars and edit counts and noticeboard reputation, a hierarchy legible and meaningful only to other people inside the same world. And it is, for many of the people who hold it, one of the few arenas in life where their actions produce large and immediate effects on others. That is not a description of damaged individuals. That is a description of a machine, and the machine has a predictable output.
Hand almost anyone power that is real but unaccountable, status-conferring but only within a closed circle, consequence-free in their outside life but enormous inside the game, and exercised mostly against strangers they will never have to face — and you will, reliably, get some fraction of them using it the way this essay describes. Not because they are bad. Because the structure has quietly removed every brake that would normally restrain the use of power: the social cost of being seen to be cruel, the reciprocal vulnerability of the person you’re being cruel to, the possibility of being held to account by someone who outranks you. Remove the brakes and the car rolls downhill. It does not require a villain at the wheel. It requires only a hill and the absence of brakes, and Wikipedia’s governance has built, with the best intentions, a very long hill.
This is why the “they’re just losers” theory is not only unkind but actively dangerous: it locates the problem in a few bad apples, which implies the solution is to find and remove the bad apples, which the system is already nominally set up to do and which changes nothing, because the next person into the same incentive structure inherits the same incentives. The barrel is the problem. The hill is the problem. You will not fix an autoimmune disease by apologizing to individual white blood cells.
The case for the defense, made as well as I can make it
I would be doing exactly the thing I’m accusing the wikilawyers of — using rhetoric to win rather than to clarify — if I left it there, with the prosecution rested and the defense unheard. So let me now make the moderators’ case as strongly as I can, because it is strong, and because any version of this critique that can’t survive it isn’t worth writing.
The moderators would tell you that you have no idea what they hold back. And they would be right. The picture I have painted — the eager newcomer crushed by a territorial veteran — is real, but it is a curated slice. For every innocent newcomer fixing a founding year, the people on the front line are also fielding a relentless, exhausting, infinite tide of the genuinely malicious and the genuinely incompetent: the spammers, the marketing departments laundering their clients’ reputations, the political operatives scrubbing inconvenient facts, the conspiracy theorists who will not stop, the hoaxers, the libelers writing vicious falsehoods into the biographies of real living people, the paid editors who lie about being paid, the sockpuppets — single individuals operating armies of fake accounts to manufacture the appearance of consensus. The moderator who reverts your good edit in ninety seconds with a curt summary has, very often, just reverted forty bad ones, and the curtness you experienced as contempt is, from the inside, simply the pace required to survive the volume. They cannot afford to write you a warm personalized note. There are nine hundred more edits in the queue and most of them are garbage and they are doing this for free, on their own time, because no one else will.
And the assumption of good faith, which sounds so humane in the abstract, runs into a hard wall of experience. Assume good faith is a wonderful principle for your first month and a corrosive one for your fifth year, because by your fifth year you have been lied to ten thousand times. You have watched the same playbook run a hundred times: the friendly newcomer who is actually a banned user back under a new name, the reasonable-sounding source request that is actually the opening move of a coordinated POV campaign, the polite question that is bait. The veteran’s reflexive suspicion that reads, to a newcomer, as paranoid contempt is, to the veteran, simply learning. It is pattern recognition paid for in years of being burned. Tell a frontline immune cell to assume the next foreign body is harmless and you have not made it kinder; you have made it useless.
The rules I’ve mocked as a cudgel are also the only thing standing between the encyclopedia and chaos. Notability standards keep Wikipedia from drowning in autobiographies and garage bands and small-business advertisements. Verifiability and reliable-sourcing rules are the entire reason the thing is trustworthy at all — the reason you can look something up and believe it, the reason it is not just a billion-page comment section. The shorthand that gets wikilawyered is also what lets thousands of strangers coordinate at planetary scale without descending into permanent war. Take the rules away to be nicer to newcomers and you do not get a kinder Wikipedia. You get a worse encyclopedia, fast, and then no encyclopedia at all.
All of this is true. I believe every word of it. And here is the thing that makes the problem genuinely hard rather than a simple matter of good guys and bad guys: the defense and the prosecution are both correct. The behaviors that protect the encyclopedia from its enemies are very often the same behaviors, applied to the wrong target, that drive away its friends. The suspicion that catches the sockpuppet bites the newcomer. The speed that clears the vandal queue feels like contempt to the person who came to help. The rules-fluency that defeats the POV-pusher humiliates the expert. There is no clean line where vigilance ends and cruelty begins, no setting on the dial where the immune system attacks every invader and no healthy cell. That is exactly why the autoimmune turn is so insidious and so hard to reverse: the symptom and the cure are made of the same material. You cannot simply tell the moderators to stop, because the thing you’d be telling them to stop doing is, ninety percent of the time, the thing keeping the lights on.
Which means the answer is not less vigilance. The answer is better-aimed vigilance — and that is a much harder, much less satisfying thing to ask for than a villain to depose.
What it costs, and who pays
Before the prescriptions, the bill, because the bill is the reason any of this matters. If the autoimmune turn merely bruised some feelings, it would be a manners problem and not worth six thousand words. It is worth the words because of what it does to the encyclopedia itself.
The first cost is the obvious one: people leave, and they take their knowledge with them. Every newcomer ground down into departure is not just a lost editor but a lost body of knowledge — the things they knew that no one else who stayed happens to know. And the loss is not random. Attrition selects. The people most likely to endure years of procedural combat, talk-page sieges, and noticeboard theater are people with a high tolerance for exactly that: combative, rules-oriented, contentious, and — this is the part that should worry anyone who cares about the project — demographically narrow. The people most likely to leave are the ones with less time to spare, less appetite for conflict, and less prior membership in the culture that already dominates. Over years, this is a filter, and a filter has an output, and the output is a community that looks less and less like the world it claims to be documenting.
Wikipedia knows this about itself, to its credit. The community has long acknowledged its own systemic biases — there are entire volunteer projects devoted to countering them — and has been candid for years about the lopsidedness of who actually edits: overwhelmingly skewed in ways the project openly frets about. You can see the consequences in the coverage itself, in the strange unevenness of what the encyclopedia knows. Astonishing depth on the things the dominant editor culture finds interesting; thin, contested, or deleted-as-non-notable coverage of the things it doesn’t. The topics that matter to communities underrepresented in the editor base are exactly the topics most likely to be challenged for “notability,” fought over, and quietly removed — not usually out of malice, but because notability is judged by people, and people judge importance by what looks important to them, and the people doing the judging are not a cross-section of humanity. They are the ones who survived the filter.
So the homogenization is not a vague worry about vibes. It is a mechanism with an output you can read directly off the encyclopedia’s table of contents. A platform whose entire moral claim is to gather the sum of all human knowledge slowly becomes a record of the knowledge held by the specific, narrow, self-selected population willing to wage years of low-grade war to keep their edits live. The sum shrinks toward the sum of what the survivors know. And the cruelest part is that it happens invisibly, edit by reverted edit, with no villain and no announcement, each individual loss too small to notice and the aggregate large enough to bend the shape of recorded human knowledge.
The second cost is to trust, and trust is the only thing Wikipedia actually has. It owns no printing press, commands no authority, holds no credential. It is believed because it is seen to be fair — open, correctable, neutral, a place where the truth wins on the merits and anyone can check the work. Every time the process visibly rewards the entrenched over the correct, every time a good edit dies to a territorial revert, every time a newcomer is bitten and the biter walks, a little of that trust leaks out. Not all at once. But the promise on the front door — anyone can edit — is a promissory note, and a note that bounces often enough stops being money. The danger is not that Wikipedia becomes wrong. The danger is that it becomes a club that has stopped admitting members while still advertising that the door is open, and that the gap between the advertisement and the reality eventually becomes the only thing anyone can see.
Reclaiming the door
So what is to be done — not in the language of a corporate slide deck, which the encyclopedia does not need and would not heed, but plainly?
The first thing is the hardest, because it is cultural rather than technical: the rules that protect people have to be enforced with the same seriousness as the rules that protect content. Right now the asymmetry is total. Bad sourcing is policed in seconds; biting a newcomer is policed never. As long as that holds, every other reform is decoration, because the incentive structure will simply route around it. “Do not bite the newcomers” cannot be a poster on the wall that everyone walks past. It has to be a thing that, when violated flagrantly and repeatedly, costs the violator something — the way violating a sourcing rule costs you. Until the civility floor is load-bearing, it isn’t a floor. It’s a mural.
The second is transparency in the most literal sense: explanation. The single cruelest thing about the ninety-second revert is not the revert — sometimes the revert is correct — it is the silence. A revert with no human explanation tells the newcomer nothing except that they are unwelcome. A revert with one sentence of plain language — the founding year needs a source we can check; here’s how to add one — converts the identical action from an act of gatekeeping into an act of mentorship, at a cost of about fifteen seconds. The encyclopedia should make the explained revert the norm and the silent one the exception, and it should build its tools to make explanation the path of least resistance rather than, as now, the path of most.
The third is appeal that does not route through the accused. Right now, far too often, the person you must appeal to about an abuse of power is structurally aligned with — or literally is — the person who has the power. You cannot fairly challenge an authority by petitioning that same authority, and a community serious about fairness needs avenues of recourse that are genuinely independent of the cliques whose decisions are being challenged. This is the oldest principle in the design of just institutions, and Wikipedia, for all its sophistication, has never fully solved it.
The fourth is mentorship as a first response rather than a remedial one — a deliberate cultural shift from “correct the transgression” to “induct the newcomer,” from treating the unfamiliar editor as a defendant to treating them as an apprentice. Every veteran who learned the dialect was, once, someone who did not know it. The willingness to remember that, and to extend the same patience that was once extended to you, is not softness. It is the literal mechanism by which a community reproduces itself, and a community that stops doing it is a community that has decided to die, slowly, of its own competence.
And the last, the one that subsumes all the others, is simply this: the guardians have to remember what they are guarding for. The immune system exists for the body. The moderation exists for the encyclopedia, and the encyclopedia exists for the contributors and the readers — for the retiree fixing her town’s history, for the grad student citing a real paper, for the stranger somewhere right now correcting a founding year in good faith. When the defense of the thing becomes indistinguishable from an attack on the people the thing is for, the defense has lost the plot, no matter how vigorously it is performed, no matter how many bad edits it correctly catches along the way. Working hard is not the same as working right. An autoimmune system is working very hard indeed.
I keep returning, in my mind, to that newcomer at the start — the one fixing a sentence, feeling for a moment like a participant in something enormous, before a stranger they will never meet undoes the whole thing without a word. I want to be fair to the stranger. The stranger has probably just reverted forty pieces of garbage and is tired and is unpaid and is, in the largest sense, on the right side — standing between the cathedral and the latrine wall, doing the thankless necessary work that keeps the miracle from collapsing into a comment section. The stranger is not the villain. There is no villain. That is the whole unbearable point.
But the newcomer is not a threat, either. The newcomer is the encyclopedia’s actual future, the next body of knowledge, the next decade of contributors, the living tissue the entire immune system exists to protect — and the encyclopedia just taught them, in their very first interaction, that the door marked anyone can edit opens onto a room where they are presumed guilty and given no reason. Some of them will fight through that and become editors anyway, and those few will be tougher and more contentious and more like the people who are already there, which is part of how the filter does its quiet work. Most of them will simply close the tab.
The encyclopedia will not notice. There is no log entry for it. The page will stabilize, the keeper’s version will stand, and the immune system will record one more success. And the sum of all human knowledge will be, by one quiet increment, a little smaller than it could have been — guarded so well, and so badly, by people who can no longer tell the difference between defending the thing and devouring it.
The cathedral still stands. It is still one of the best things we have made. I want it to stay that way badly enough to say the part that sounds ungrateful: a guardian who has forgotten what he guards is not a guardian. He is just the last person in the building, turning away everyone who comes to help him keep it standing, certain to his bones that he is doing his job — and he is, he is doing it ferociously, right up until the day there is no one left inside but him.