Track changes — v1.0.0...HEAD

The volunteers remain. About 250,000 editors still maintain Wikipedia every month, still fixing errors, adding citations, arguing on talk pages about neutrality and sourcing.\autocite{wikimediafoundation:wikipedia-celebrates-25years} They still believe their contributions matter, and their choice will decide the outcome.

If they conclude the promise was broken and drift away, the database will persist but the community will scatter. Another machine will go dark. If instead they organize, protest, fork, or simply refuse to accept that extraction is inevitable, the Foundation will have to answer. The Spanish editors who forked in 2002 forced Jimmy Wales to abandon his advertising plans.\autocite{cpov2011spanishfork} Community pressure has bent Wikipedia's course from the start. The volunteers are not powerless. They have not yet chosen to use their power.that power at scale.

\scenebreak

All web sources have been archived using the Wayback Machine. Wikipedia citations include revision IDs to ensure verifiability even as articles change.

Mailing-list quotations were verified against a local corpus of the movement's public list archives: the complete pipermail mbox files for eight Wikimedia lists from 2001 through the March 2021 platform migration, nearly half a million messages, plus mirrored archives of the surviving lists from 2021 to July 2026 and the short-lived Sifter-L list, recovered from Wayback Machine captures. Claims about what the lists did \emph{not} discuss, such as the near-silence on GPT-3 before ChatGPT's launch, rest on full-text searches of that corpus.

So that readers can check the record for themselves, the combined corpus---489,491 messages in a single normalized file, with each message's list, date, author, subject, and source archive---is available for download at \texttt{commonscaptured.com}, alongside a complete change record for this printing. The underlying archives are public; the download saves only the work of reassembling them.

\section*{The FEC Donation Analysis}

My own proximity to these events is disclosed in ``A Personal Note,'' which precedes this appendix.

A note on participation. Larry Sanger corresponded with me during the preparation of the second printing, sat for an interview, and reviewed the passages that quote him; his testimony is cited as such and checked against the contemporaneous record wherever that record survives.

In January 2000, Wales hired him to develop and run Nupedia. Sanger arrived in San Diego the following month to begin work.\autocite{Sanger2005Memoir}

The philosophical tension between them was present from the start. Wales believed that good processes could substitute for credentials. Sanger believed that credentials signaled reliability precisely because knowledge requires justification, and justification requires expertise. Wales trusted emergence. Sanger trusted verification.\autocite{Sanger2005Memoir}verification, a difference of degree that would later harden into myth.\autocite{Sanger2005Memoir}

For Nupedia, Sanger designed a process that reflected his training. If knowledge required justification, then articles would need to be justified---by experts, through review, at every stage. The result was a seven-step gauntlet that looked exactly like academic peer review because, in Sanger's mind, academic peer review was how knowledge got validated.\autocite{Sanger2005Memoir}

Applied to encyclopedias, the idea was radical. Maybe the problem was not finding enough experts. Maybe the problem was relying on experts at all. Knowledge might emerge from the coordination of thousands of non-experts, each contributing a small piece, with the aggregate exceeding what any committee could produce.

This was heresy toTo Sanger's epistemology. Knowledge requires justification. JustificationNupedia-trained instincts, this ran against the grain: knowledge requires justification, and justification usually comes from expertise. Without credentials, how could you trust what someone wrote?But the two men were never as far apart as the later mythology made them. Sanger believed in distributed knowledge and open processes too, and would spend the next year championing the wiki harder than anyone. Where they parted was narrow: expertise, he thought, deserved some role inside the process.\autocite{SangerEdge2007,SangerInterview2026}

Hayek's model suggested a different kind of trust: trust in process rather than credentials. If enough people contributed, and if good processes corrected errors, then the quality of any single contributor mattered less than the scale of participation.

How much of that theory was present at the creation, and how much was fitted to it afterward, is disputed. Wales told the Hayek story often in later years. Sanger, who watched from inside, read it as a retrofit: Wales the Bomis taxonomy-builder was convinced by evidence, not theory, when Nupedia crawled and the wiki took off.\autocite{SangerInterview2026} Either way, the system that emerged behaved as if the theory were true.

The question was whether such a process could exist. In January 2001, Sanger stumbled across one.

By January 2003, the English edition alone had passed 100,000 articles.\autocite{en:History_of_Wikipedia} The experiment had become the project.

The Nupedia Advisory Board had grasped the stakes correctly. The wiki was not just a technical change but a philosophical break. Nupedia treated knowledge as something experts produce and gatekeepers validate.reviewers confirm. Wikipedia treated it as something that emerges from process, from the contributions of thousands of people, most anonymous, none credentialed. The Board saw thatbelieved these were incompatible visions, and that a wiki would undermine Nupedia's whole premise. They misjudged only which model would win.

What the Board did not anticipate was that the internet had changed the economics of knowledge production. Nupedia needed experts who would work for free despite no professional incentive to do so. Wikipedia tapped a different pool: people who wanted to build something larger than themselves, who had passion for their subjects, who found satisfaction in a shared resource. These people existed in the millions. Academics with PhDs existed in the thousands. Scale beat credentials, and almost no one saw it coming, Sanger included.

On September 26, 2003, Nupedia's servers were shut down for good.\autocite{en:Nupedia} By then, Wikipedia had more than 150,000 articles in English alone, created by contributors whose only credential was the willingness to edit.\autocite{en:History_of_Wikipedia}

Sanger had left the projects 18 months earlier,payroll in March 2002.\autocite{SangerNoLongerParticipating} But2002 and cut ties entirely early in 2003, after a last private appeal to Wales, to rein in the philosophical battle was already lost bytrolls and give experts some low-key role, failed to move him.\autocite{SangerNoLongerParticipating,SangerCorrespondence2026} By then the time he left. Wikipedia had chosen Wales's model over Sanger's.outcome was visible either way. The communitywiki had voted with their keystrokes, and they voted for velocity over verification.outrun the review process. Contributors chose velocity. Verification found few volunteers.

The original idea had been an encyclopedia written by experts, for everyone, with the wiki serving as its feeder system. Instead, the feeder ate the host.

One consequence emerged more slowly. Wikipedia's anti-credentialist philosophy said nothing about who would staff the Foundation built to steward it. That would turn on hiring networks, geographic sorting, and cultural fit, and those forces would produce a homogeneity of their own---one that would, decades later, become measurable in the public record. An institution composed of like-minded people can move quickly and agree easily. It can also lose the capacity to recognize, when extraction finally arrives, that the people arriving to extract are not partners.

The founding tension, between openness and control, between process and credentials, was never resolved. The Spanish-language editors who broke away in 2002, fearing commercial control, fractured over these questions first.\autocite{cpov2011spanishfork} Popular memory reads Sanger's departure madeas thephilosophical split permanent.made permanent; Sanger himself denies there were ever two competing models, only a role for experts that never got built.\autocite{SangerCorrespondence2026} And when AI companies began taking Wikipedia wholesale, the openness its founders celebrated offered no resistance. The license did, but only if someone chose to enforce it.

The question was no longer whether the open model could work. It had worked spectacularly. The question was what it would cost, and who would pay.

"I asked myself 'why are we working for a dot-com?'" he recalled in a 2011 interview. "I asked for Wikipedia to be changed to a dot-org."\autocite{cpov2011spanishfork}

He objected to Larry Sanger's role as editor-in-chief: "I wanted the Big Brother out. Larry Sanger was against the nature of the project itself. None of us felt comfortable with such a figure."\autocite{cpov2011spanishfork} That was one participant's reading. The role Sanger says he wanted for experts was narrower, "not a supervisory or editorial role" but an article-rating layer, "one that might not even be part of Wikipedia itself."\autocite{SangerCorrespondence2026}

He pushed for autonomy: "We did not want to be seen as mere translations of the American version... I was told so many times to translate from the main wiki, and my response was always the same: We are not a translation of the American Wikipedia!"\autocite{cpov2011spanishfork}

\section*{The Fork}

In late 2001, the internet bubble collapsed. Bomis lost major contracts. Revenue dried up.\autocite{sangerOnAdvertisement} By December, Tim Shell, a Bomis co-founder, had told Larry Sanger hehis paid position would be laid off.cut. The layoff took effect in February 2002. Wikipedia was costing money Bomis no longer had.\autocite{Sanger2005Memoir,sangerOnAdvertisement}

On February 2, 2002, Jimmy Wales posted to the Wikipedia-L mailing list: "However, with the ongoing hard times in the Internet economy, we do anticipate adding some forms of advertising to the site in the near future."\autocite{walesWikipediaL2002}

\section*{Wales Responds}

On March 1, 2002, Larry Sanger resigned as Wikipedia's paid chief organizer.organizer; he laid the role down entirely but stayed on the wiki, at a fraction of his old pace.\autocite{SangerInterview2026} The same day, Wales sent an email to Wikipedia-L with a significant reversal:

"With the resignation of Larry, there is a much less pressing need for funds. Therefore, all plans to put advertising of any kind on the wikipedia is called off for now."\autocite{walesWikipediaLMar2002}

The mailing list archives contradicted Wales's later account.\autocite{walesWikipediaL2002} He had announced advertising plans. The community had revolted. He had reversed course. These were documented facts.

But the dispute revealed something deeper. On the advertising question, the contemporaneous record backs Sanger's account over Wales's later telling.\autocite{walesWikipediaL2002,sangerOnAdvertisement} The founders themselves couldn't agree on what they had intended.\autocite{sangerOnAdvertisement} Nodeeper failure was structural: no governancestructure existed to settle such questions.questions, or to stop a founder from rewriting their history. Everything depended on the memory and goodwill of the people who controlled the servers.

The Spanish contributors fought about advertising because advertising was visible and concrete. The underlying problem remained even after the Foundation was created: no governance mechanism constrained central authority.

From Wales's perspective, and the official Wikimedia Foundation history, the fork was a misunderstanding resolved. Wikipedia was always meant to be ad-free. The February 2002 announcement was an aberration caused by financial pressure, not a statement of intent. The Foundation's creation reflected natural evolution, not forced concession. The project "stayed the course."\autocite{sangerOnAdvertisement}

Both accounts hold a piece ofOn the ads themselves, the archives side with the truth,forkers and with Sanger. What both tellings miss is what Enyedy had actually seen. The fork was not only about advertising. It exposed a pattern that would define Wikipedia's next two decades. A shared resource stays stable only while its contributors believe they control it. The moment a central authority takes it for profit, the social contract collapses.

The forkers were right about the threat. They were wrong only about its form. They feared advertising. They could not have imagined the Foundation one day selling bulk access to its content for AI training.\autocite{techcrunch:wikimedia-foundation-announces-new-ai-partnerships-with-amaz} Nor could they anticipate a subtler capture: the institution created to protect against commercialization would come to be led by people who saw tech companies as natural allies, not adversaries. They won concessions against banner ads, the visible symbol of commercialization.\autocite{walesWikipediaLMar2002} They established no protection against extraction through training data.

\chapterquote{As a community, Wikipedia lacks the habit or tradition of respect for expertise.}{Larry Sanger, 2004}

\person{Larry Sanger} explained his departure twice, and the two explanations did not match.left Wikipedia twice. The record long remembered it as once.

HisThe first departure was formal and financial. His resignation, posted on March 1, 2002, said all the right things: no loss of faith in the project, a relationship with Jimmy Wales and Bomis that "remains friendly," a departure that was purely financial.an exit forced by money alone. The dot-com bubble had burst, and Bomis could no longer afford his salary.\autocite{MyResignationSanger,SangerNoLongerParticipating}salary.\autocite{MyResignationSanger}

This was true as farWith it he laid down the organizer's role entirely, and by his own account voluntarily: he could have kept the title as an unpaid volunteer, and chose not to.\autocite{SangerInterview2026} Newly married, with a living to earn, he spent most of the year that followed focused elsewhere, and the project grew and changed without him.\autocite{SangerInterview2026} The second departure came roughly a year later, and it went. But the fuller confession, written months later onwas the real break. On February 27, 2003, he posted an essay titled "Why I Am No Longer Participating in the Wikipedia Project" to Meta-Wiki, the community's side site for discussing the project itself, told a different story:itself:

"I must now confess, when I had to quit as 'organizer' of Wikipedia at the end of February 2002, that it was a relief."\autocite{SangerNoLongerParticipating}

A relief. The only paid Wikipedia employee. The man who named the project, designed its policies, served as its chief organizer for its entire first year.\autocite{Sanger2005Memoir} And leaving felt like escape.

Because the 2003 essay looked back at the 2002 layoff, later histories collapsed the two departures into one. The distance between them is where the real story sits.

"Very much of my time was spent in defending myself from such attacks," he continued. The attacks came from within, from other Wikipedians who opposed his vision of how the project should work. Certain critics, he wrote, had mounted "a concerted effort to criticize and undermine virtually every 'management'" action he took.\autocite{SangerNoLongerParticipating}

The wiki structure worked against him. "Critics in a wikiwiki environment cannot be silenced... no criticism, no matter how false or misleading, can be erased without an accusation of censorship." Every dispute escalated. Every editorial decision became a referendum on his authority. By the time Bomis informed him the funding would end, the poisonous atmosphere had already driven him toThat atmosphere, he wrote in 2003, was what made the brinkend of quitting.\autocite{SangerNoLongerParticipating}his job feel like escape. Whether it drove either departure is a separate question, and Sanger's answer, for the first one, was a flat no.\autocite{SangerNoLongerParticipating,SangerCorrespondence2026}

What happened in thoseacross that long final months?year? What tension could make the creator of Wikipedia's policies desperaterelieved to leave his own creation?creation, and then pull him back, near the year's end, for one last fight over its direction?

\section*{The Partnership}

In January 2001, when Wikipedia launched, \person{Larry Sanger} was its only paid employee. Jimmy Wales provided funding through Bomis and made high-level decisions, but Sanger was the one building the project day by day.\autocite{lih2009wikipedia} He wrote or commissioned the first articles. He formulated the Neutral Point of View policy that would become Wikipedia's philosophical foundation. He moderated disputes, recruited contributors, and shaped the community's norms.\autocite{Sanger2005Memoir}

The press releases from those early months credited both men as founders.\autocite{SangerRole} In September 2001, the New York Times reported that Sanger had "founded Wikipedia with Mr. Wales," a characterization Wales disputed for years afterward.\autocite{NYTMeyers2001,wiki:Jimmy_Wales} The relationship appeared collaborative. Two men with different skills, building something together. The shared dream, as Sanger later put it, was to beat Britannica.\autocite{SangerInterview2026}

But a tension in emphasis between the two men never resolved. It was not a clash of two grand models. Both believed in distributed knowledge and emergent order; Sanger was the wiki's most tireless evangelist, and he would later write that experts should share responsibility for the world's knowledge with the general public, not hold it alone.\autocite{SangerEdge2007} The difference was narrower. Sanger thought expertise deserved some standing role in quality control, a mainstream assumption in 2001. Wales's Wikipedia developed no such role at all.\autocite{Sanger2004AntiElitism,SangerInterview2026}

But the philosophical tension between the two men had never resolved. Wales believed in distributed knowledge, emergent order, process over credentials.\autocite{Reagle2010} Sanger believed expertise should have standing, that credentials signaled reliability, that quality required some form of gatekeeping.\autocite{Sanger2004AntiElitism}Wikipedia had chosen to be open. Sanger kept trying to impose structure onbuild order into that openness.openness---in his eyes the only thing that could preserve it. "There can be no true openness," he later argued, "if bad actors are constantly driving away some of the most active, knowledgeable people in the project. It's either the trolls or these other people."\autocite{SangerCorrespondence2026}

For a while, both visions coexisted. Then they didn't.

The attacks weren't coordinated, but they shared a common theme: Sanger was too controlling. His attempts to maintain standards were authoritarianism. His editorial judgments were elitism. Every time he tried to enforce quality, someone accused him of violating Wikipedia's open culture.\autocite{SangerNoLongerParticipating}

The structural problem was that Wikipedia had no mechanism to resolve such disputes. In a traditional organization, Sanger's titlerole as editor-in-chiefchief organizer would have settled things.things.\footnote{On Nupedia, Sanger was editor-in-chief; on Wikipedia he styled himself "chief organizer"; the press applied both titles to both projects. This book uses chief organizer for his Wikipedia role.} In Wikipedia's radical flatness, titles meant nothing. Anyone could criticize anything. And in a wiki environment, criticism couldn't be removed without triggering accusations of censorship.\autocite{Reagle2010,Sanger2005Memoir}

"It's now generally the case that if I attempt to do or say anything that is very controversial---certainly anything regarding Wikipedia policy and politics---I am either ignored or, more often, vociferously denounced," Sanger wrote.\autocite{SangerNoLongerParticipating} TheIn the essay's telling, the community he had helped build was rejecting him. Sanger himself would later put it differently. The antagonists he had in mind were figures who were "not especially important members of the community, and their complaints had nothing to do with my departure."\autocite{SangerCorrespondence2026,CunctatorAdminsLMS2001} What failed was not the community's regard for him but the failure to protect the project from the chaos and malevolence that emerged.

\section*{Troll Problem}

"Trolls" in early internet parlance were users who disrupted communities for entertainment. They started fights, inserted false information, and wasted contributors' time with bad-faith arguments.\autocite{Donath1999} A traditional encyclopedia could simply ban them. Wikipedia's commitment to openness made that much harder.

Sanger tried to address the problem. Heproposed giving experts more authority. He suggested mechanisms to identify and removerestrain troublemakers. For experts, the role he wanted was, by his own account, modest: not supervision, not editorial control, but an article-rating effort he called the "sifter" project, which might not even have lived inside Wikipedia.\autocite{SangerCorrespondence2026,SangerInterview2026} Each proposal met the same objection: you're betraying Wikipedia's principles. The project's most active contributors were "decidedly anti-elitist," he observed.\autocite{Sanger2004AntiElitism} They saw his attempts to impose structure as attacks on the open model that was working.

Sanger's later name for the hard core of that opposition was "wiki anarchists": people opposed not to this or that authority but to any exercise of power at all, even the removal of a bad actor by democratic process.\autocite{SangerInterview2026}

The irony was bitter. Sanger had designed many of Wikipedia's policies. Now those policies were being used against him.

In late 2001, the internet economy collapsed. Bomis, which had been funding Wikipedia through advertising revenue, lost major contracts.\autocite{SangerNoLongerParticipating} In December, Bomis told Sanger it could no longer afford his salary. He was one of several employees the crash swept out.\autocite{Sanger2005Memoir}

Wales promised another push to sell advertising on Wikipedia to keep Sanger on staff.\autocite{sangerOnAdvertisement} That attempt led to the February 2002 advertising fight that drove Spanish-language editors to break away and start their own encyclopedia.\autocite{cpov2011spanishfork} But for Sanger personally,On Sanger's own account, this first exit was exactly what the financial newsresignation note said: a layoff, because the money was secondary. The community had already made his position untenable.gone, and for no other reason.\autocite{SangerCorrespondence2026,SangerInterview2026}

When the public resignation came on March 1, 2002, the note tried to hold the project together. "My departure should not be taken as a reflection on Wikipedia, or you," Sanger wrote. "It still might succeed brilliantly."\autocite{MyResignationSanger} That same day, Wales emailed Wikipedia-L and called off the advertising plans that had provoked the breakaway.\autocite{walesWikipediaLMar2002} The forkers had won that fight. Sanger did not return.The paid position never came back.

\section*{Who Decides}\section*{Community Response}

The deeper problem outlastedHow did the resignation. Even as a volunteer, Sanger found he could not contribute "either with regard to policy and (collective) management orcommunity react to losing its founding organizer? In the improvement of articles, without myself feeling...responses that at any thing I say, I will be forced into a completely ridiculous, unseemly, undignified flamefest."\autocite{SangerNoLongerParticipating}survive publicly, with muted relief.

He had tried to institute "a policy\person{Erik Moeller}, later the Wikimedia Foundation's Deputy Director, wrote on Kuro5hin: "Larry Sanger played a crucial role in the first year of respectingWikipedia and deferring politely to experts."\autocite{SangerNoLongerParticipating} It failed. The community would not accept expertise ashas helped establish a basisshared philosophy for authority. The only authorityits users. At the same time, many felt that his authoritative tone was consensus,unnecessary and consensus had turned against him.in contradiction to Wikipedia's goals. This may well be the right time to abandon the concept of 'editorship' altogether\ldots"\autocite{MoellerKuro5hin2002} The transition from editorial leadership to self-regulation was framed as progress, not loss.

His departure exposed the deepest question facing Wikipedia as a commons. WhoOthers on Meatball Wiki noted: "The project now no longer hasthe authority to make decisions? In a traditional encyclopedia, the editors decide. In a market, the buyer decides. In a commons,leader (or, put another way, everyone affected should in theory haveis a voice. But "everyone" was millions of anonymous people who contributed once and vanished.leader now)."\autocite{MeatballSangerDeparture}

In these public responses, at least, relief outweighed grief. Sanger disputed that they spoke for the community: the celebrants, he maintained, were the faction he had triedbeen fighting, and quieter members committed to make expertisean orderly open project were not glad to see him go.\autocite{SangerInterview2026} But the deciding principle. The community rejected it.\autocite{SangerNoLongerParticipating} The rejection wasvisible verdict stood. Wikipedia would be governed by process, not anti-intellectual. It was fiercely democratic, and democracy in a commons is messy.authority. By consensus, not expertise. By crowds, not editors.

The departure was not just from a job. It was from a community that had decided it did not want what he offered.\section*{The Long Goodbye}

\section*{Community Response}The resignation thinned his connection to the project without severing it.

How didFor months he was mostly a bystander. His work on the wiki itself collapsed, by his own estimate to perhaps two percent of the year before, and his standing in the community reactfell with it. That was less a verdict than a consequence: he had chosen to losing its founding organizer? With muted relief.step back.\autocite{SangerInterview2026} In April he surfaced on the mailing list with essays on the project's troll problem, proposing ways to contain "trolls and other nuisances" without breaking the open model.\autocite{SangerTrolls2002} Then, for a long stretch, little.

\person{Erik Moeller}, later the Wikimedia Foundation's Deputy Director, wrote on Kuro5hin: "Larry Sanger played a crucial roleThe re-engagement came in the first yearfall. On the project's new English-language list, his November subjects read like a bill of Wikipedia and has helped establishparticulars: "What we need." "Wikipedia is not a shared philosophy for its users... This may well becreche." "Clean up our mess!"\autocite{SangerWikiENNov2002} And he was not shouting into the void. "I think he's winning," Wales told the right time to abandonlist that November, "and has the conceptsupport of 'editorship' altogether."\autocite{MoellerKuro5hin2002} The transition from editorial leadership to self-regulation was framed as progress,almost every regular---in goal, if not loss.in all aspects of methodology."\autocite{WalesSupportSangerNov2002}

OthersWhat he wanted most was not a procedural fight. It was the sifter, announced on Meatball Wiki noted:November 5 as a "peer review project" for Wikipedia: subject experts "picking and posting" the best articles, a task he promised would be "exceedingly simple," producing "a \emph{subset}, not a \emph{fork}, of Wikipedia articles."\autocite{SangerSifterAnnounce2002} The proposal's own terms promised to "leave Wikipedia alone completely."\autocite{SifterLInfoPage2002} "The project now no longer has a leader (or,\emph{whole purpose}," he told the project's mailing list, "is to give the world a body of articles that alleged experts have approved."\autocite{SifterLArchive2002}

For a month the sifter was a working group, not a wish. Its list, Sifter-L, drew some fifteen participants; Magnus Manske wrote software for it; Wales was copied on the discussion.\autocite{SifterLArchive2002} The archive's last message is dated December 4, 2002.\autocite{SifterLArchive2002,SifterProjectMeta}

It came to a head in December, in a fight over whether to moderate the mailing lists themselves. The moderation question mattered less to him than the sifter, but it was the fight that surfaced in public. The same days, in a content dispute over the Dark Ages, the standing of expertise itself was put plainly. "Your alleged expertise is not an argument," Erik Moeller wrote to a medievalist on the list. "It may be an argument in Larry's world, but it is not here."\autocite{SangerModerationDec2002b} Sanger joined that thread the same day, quoting the line back at Moeller; his own name had become the shorthand for the worldview being dismissed.

The same Erik Moeller had greeted his resignation in March as the right moment to abandon editorship altogether. When another way, everyonedebater threatened to leave, Sanger replied: "'I'll leave if you do' isn't very much of an argument, though it might have an effect."\autocite{SangerModerationDec2002a}

On December 12, 2002, he posted a note titled "Advice I'm following: take a break." "I'm going to take a break from the lists of at least one week," he wrote. "If a moderated list is not to be---which is a leader now)."\autocite{MeatballSangerDeparture}damn shame, because we *REALLY* need one..."\autocite{SangerTakeABreakDec2002}

In theseThe break became permanent. The archive records no further posts from him that winter. The list's daily reality that month argued his case without him. A week after his withdrawal, a thread titled "Wikipedia is under attack" tracked a vandal hopping across IP addresses: "We need to do something quick."\autocite{ZoeUnderAttackDec2002} When one participant proposed requiring log-ins for proxy users, Wales demurred: "I'd prefer if we didn't do that," suggesting rate limits instead.\autocite{WalesUnderAttackDec2002}

The final act came that winter, around the turn of 2003: a private email to Wales, sent quietly to avoid an unnecessary public responses, at least, relief outweighed grief.row, which is one reason the archives hold no trace of it.\autocite{Sanger2005Memoir,SangerCorrespondence2026} The anti-elitist philosophygrievance was not the trolls themselves but the man who tolerated them: he was, as he later put it, "extremely frustrated that Jimmy Wales was allowing trolls to take over the place."\autocite{SangerCorrespondence2026}

The demand had two parts. Wales should personally crack down on the problem users. And he should make the project more welcoming to experts, who "had been leaving in droves for the last 18 months." Otherwise Sanger would break with the project altogether.\autocite{Sanger2005Memoir,SangerCorrespondence2026} In the memoir he was careful to say he "did not mean to issue an ultimatum." Two decades later, he called it exactly that. Either way, the demand went unmet, and he kept his word: "That's when I stopped editing Wikipedia."\autocite{SangerCorrespondence2026}

The written record cannot date it more precisely. Sanger also recalls, two decades on, that the following spring Wales told him the Bomis server had struggled againstbeen "hacked."\autocite{SangerCorrespondence2026} The project's public logs for monthsspring 2003 show hardware failures, a server migration, and, in late May, a hastily disabled bug that had won. Wikipediabeen exposing user passwords and email fields to administrator queries.\autocite{VibberSecurityMay2003} They record no break-in, but a loss confined to private accounts would be governedleave no public trace; the silence neither confirms nor contradicts the recollection. What survives is the public arc: a moderation fight that failed in December, a farewell note, and silence.\autocite{WikipediaMilestones2003,WikimediaNews2003}

On February 27, 2003, Sanger posted the essay explaining why he was no longer participating, and asked an administrator to lock it.\autocite{SangerNoLongerParticipating} The second departure was complete.

\section*{Who Decides}

The deeper problem had outlasted the resignation by process,a year. Even from the sidelines, Sanger found he could not contribute "either with regard to policy and (collective) management or to the improvement of articles, without myself feeling... that at any thing I say, I will be forced into a completely ridiculous, unseemly, undignified flamefest."\autocite{SangerNoLongerParticipating}

He had tried to institute what he later called "the policy of respecting and deferring politely to experts."\autocite{Sanger2004AntiElitism} It failed; "one of my failures," he counted it.\autocite{Sanger2004AntiElitism} The community would not accept expertise as a basis for authority. By consensus,The only authority was ``consensus,'' and what passed for consensus ran against him.

His departure exposed the deepest question facing Wikipedia as a commons. Who has the authority to make decisions? In a traditional encyclopedia, the editors decide. In a market, the buyer decides. In a commons, everyone affected should in theory have a voice. But "everyone" was millions of anonymous people who contributed once and vanished.

Sanger had tried to secure a standing role for expertise, even a modest one. The project refused it.\autocite{SangerNoLongerParticipating} Was the refusal anti-intellectualism, a community defending a fiercely democratic way of working, or a loud few drowning out a quieter majority? The record can sustain each reading. Which voices really spoke for Wikipedia in those threads is exactly what the project had no way to decide.

The departure was not expertise. By crowds,just from a job. It was from a project that refused expertise even a modest role, and would not editors.defend the contributors being driven out.

\section*{The Divide}

Sanger made the underlying conflict explicit in December 2004, in an essay on Kuro5hin. Titled "Why Wikipedia Must Jettison Its Anti-Elitism," it articulated the critique he had been developingbuilding since his departure.\autocite{Sanger2004AntiElitism}final break with the project.\autocite{Sanger2004AntiElitism} By his own reckoning it was his first public criticism of Wikipedia, arriving two years after he stopped editing.\autocite{SangerCorrespondence2026}

Experts would avoid Wikipedia, he argued, because their credentials carried no weight. "Nearly everyone with much expertise but little patience will avoid editing Wikipedia, because they will---at least if they are editing articles on articles that are subject to any sort of controversy---be forced to defend their edits on article discussion pages against attacks by nonexperts."\autocite{Sanger2004AntiElitism}

He acknowledged that none of this would change. "I certainly do not expect Jimmy Wales to change his mind. I have known him since 1994 and he is a smart and thoughtful guy; I am sure he has thought through his support of radical openness and his (what I call) anti-elitism."\autocite{Sanger2004AntiElitism}

The founding schism wasn't a misunderstanding to be resolved. It was afundamental disagreement about how knowledge should be organized. Wikipedia had chosen one model.whether the wiki Sanger representedhad designed needed the road not taken.guardrails he wanted built into it, editorial guidance and some standing role for experts. Wikipedia proceeded without them.

His critique aimed at quality. But thesame openness he foughtwanted disciplined had a second edge: it made Wikipedia easy to copy, and it left no one charged with defending the terms of the license. The flat, anti-hierarchical model gave no one clear standing to speak for the project's value. Sanger had wantedenforceable authority. The community refused it, and the project to enforce its own norms; no such capacity was built. The question of who would defend its opennessthe commons passed to an institution that did not yet exist.

\section*{Two Readings}

How the departure was remembered split along the same line that caused it.

In the mainstream history, and in Wales's telling, Sanger's exit was a sad but ordinary casualty of the dot-com crash. Bomis lost money. Salaries were cut. Sanger left. Wikipedia continued and succeeded spectacularly.\autocite{lih2009wikipedia} That telling described the first departure and stopped there. The quiet second break did not appear in it, and Sanger himself was partly why: he had made his ultimatum privately, wanting no public row.\autocite{SangerInterview2026}

The evidence for that reading is substantial. Wikipedia grew from some 20,000 articles in early 2002 to over 350,000 by late 2004.\autocite{en:History_of_Wikipedia} The project stayed the course on its founding principles and became the world's largest encyclopedia. Sanger's later alternative, Citizendium, peaked briefly in 2008 and then faded, seeming to provewidely read as proof that an expert model could not work at scale.\autocite{CitizendiumWikipedia}

Wales treated the co-founder dispute as settled and tiresome. In 2009 he called it "silly."\autocite{WikipediaSignpost2009} In 2023 he said Sanger "doesn't get enough credit for his early work" while holding that "co-founder's not the right title."\autocite{lexfridman:jimmy-wales-transcript} In November 2025, asked about it on a German podcast, Wales walked out after 48 seconds and called it "the dumbest question in the world."\autocite{Storyboard2025} The tension stayed raw, but the trajectory looked clear. Wikipedia won. Sanger lost.

Sanger told it differently.differently, and with a cleaner line than either camp drew for him. The financial trigger obscured the real cause, whichlayoff was philosophical rejection. Wikipedia's egalitarian culture had made his role impossible, andmoney, nothing more. The real rupture came the community would not accept his vision of structured expertise. Evenfollowing winter, as a volunteer, when he found continuedconcluded that Wales would not restrain the project's disruptors and that his own participation untenable.\autocite{SangerNoLongerParticipating}had become untenable.\autocite{SangerNoLongerParticipating,SangerCorrespondence2026} The mainstream telling ends the story too early. The philosophical telling starts it too early. The surviving record fits his telling: two departures, one financial, one philosophical.

In his reading, Wikipedia's success came at the cost he had predicted. Citizendium failed, he argued, not because his model was wrong but because Wikipedia already held insurmountable network effects. On that, the evidence settles less than either side claims: one attempt, in one decade, against one entrenched incumbent. Two decades on, machine-written forks would retest the proposition (\chapref{ch:grokipedia}).

The name, the policies, a year as chief editor, the first articles written or commissioned: if that record did not make him a co-founder, he asked, what would the word even mean? He had also drafted the Neutral Point of View policy that anchors Wikipedia's editorial system to this day.\autocite{Sanger2005Memoir} The contemporaneous record backed him. Wikipedia's own first press release, issued in January 2002 while Wales ran Bomis, stated it plainly: "The founders of Wikipedia are Internet entrepreneur Jimmy Wales and philosopher Larry Sanger."\autocite{WikipediaPressReleaseJan2002} Yet the institution built to serve the community had no seat for him. The person who designed the governance was excluded from it.

The disagreement did not resolve. It went dormant. The question underneath it returned in every later conflict between the community and its institutions: who governs open knowledge, and should credentials matter?

\section*{What Followed}

After the organizer role ended in March1, 2002, and Sanger's remaining participation a year later, Wikipedia continued without formal editorial leadership. Wales became the sole public face of the project.\autocite{lih2009wikipedia} The anti-credentialist philosophy was locked in. The Neutral Point of View policy Sanger had drafted remained, and the community built Verifiability and No Original Research on its foundation the following year.\autocite{en:Wikipedia:Verifiability} His approach to enforcement did not survive.

The Spanish forkers fared no better than Sanger. They won concessions on advertising but not on structural change.\autocite{cpov2011spanishfork}

The co-founder dispute festered for decades, escalating with each round. In late 2005, Wales edited his own Wikipedia biography 18 times, seven of them altering information aboutthe changes stripping Sanger's co-founder status.credit.\autocite{CadenheadWorkbench2005,wiki:Jimmy_Wales} The next year he told the Boston Globe that Sanger's claims were "preposterous."\autocite{wiki:Jimmy_Wales} By 2009, incalling Sanger a public quarrel stokedco-founder was "preposterous."\autocite{MeheganBostonGlobe2006} In April 2009, provoked by entrepreneur Jason Calacanis,a fresh Wales interview, Sanger was callingposted an open letter to Wales's own Wikipedia talk page accusing him of "lies and distortions" about the founding. Wales a "fraud"deleted it.\autocite{SangerOpenLetter2009}

The next year Sanger reported the Foundation to the FBI, alleging that images hosted on Wikimedia Commons violated federal obscenity law. "I want to be on the record stating that this is wrong," he wrote the same day. The Foundation denied any violation, and a "liar."\autocite{wiki:Larry_Sanger}no charges came.\autocite{SangerFBIReport2010,WMFResponseFBI2010}

\section*{Long Echo}

For twenty-four years Sanger had used voice, the dissenter's option of arguing from within. He critiqued, he built alternatives, he published the Nine Theses, and in 2026 he organized a WikiProject, an editor-run working group, devoted to intellectual diversity inside Wikipedia itself. He urged his followers toward the internal debate.\autocite{Sanger404Media2026}

The answer was exit, and not the voluntary kind. An administrator blocked his account indefinitely, carrying out a decision reached on the community's public administrators' noticeboard.\autocite{Sanger404Media2026} Editors charged him with off-wiki canvassing, meaning he had rallied outside supporters to sway an internal debate.\autocite{Sanger404Media2026} They also accused him of treating the project as a battleground. The procedural breachposts at issue were on the public record, cited directly in the June 2026 noticeboard discussion; the dispute was real.over what they amounted to.\autocite{SangerANI2026}

The verdict was neither smooth nor unanimous. A first block landed before the discussion's seventy-two-hour window had run. It was lifted, then reimposed once the window closed and the consensus was tallied.\autocite{Sanger404Media2026} The loudest objection came from the other founder. ``I find the idea of any of this deserving an indef ban for Larry is ludicrous,'' Jimmy Wales wrote on his own talk page the morning of the final block, adding that editors needed to ``sit back and have a hard look at what they are saying.''\autocite{JimboWalesTalk2026} In 2002, Wales's judgment had settled Wikipedia's disputes. In 2026 it counted for one comment in the thread, and the thread went the other way.

But the timing closed an arc that had opened a quarter-century earlier. The co-founder who walked away in 2002, because he could not abide2003, when the project's openness,project would neither restrain its trolls nor make room for its experts, was in the end cast out by the project itself. He had once left on his own terms. This time the community decided for him.

\section*{Sanger Void}

WhenLarry Sanger departed in March 2002, Wikipedia lost more than an employee. It lostwas the only person who had claimed authority to makeever held editorial decisions.\autocite{SangerNoLongerParticipating}authority.\autocite{SangerNoLongerParticipating} He laid it down with the paid role, and by his own account voluntarily.\autocite{SangerInterview2026}

InOn February 25, 2002, just weeksdays before histhe resignation, Jimmy Wales had written:could still write: "Larry is the final arbiter of what the consensus is."\autocite{WalesArbiter2002} After Sanger left,By March 1 there was no arbiter. Sanger's later pushes for order and a role for experts failed, as \chapref{ch:sangers-departure} recounted, and by early 2003 he was gone for good.\autocite{SangerTakeABreakDec2002,SangerCorrespondence2026}

Wales inherited thisthe arbiter's role by default rather than design. He made decisions when asked, intervened in disputes when necessary, and served as the project's public face. But his authority was never formalized. Wikipedia had no constitution, no bylaws, no governance structure at all.\autocite{Reagle2010}

This worked, for a while. Wikipedia was still small enough that disputes could be resolved through conversation. Wales was present enough to adjudicate when conversations failed. The community shared enough values that most questions had obvious answers.\autocite{Reagle2010}

What did NPOV do? It was the first true commons governance policy. Instead of asking "What is true?" (which divides any community), NPOV asked "What do reliable sources say?" The reframe worked because it let people with incompatible worldviews work on the same articles. An atheist and a believer could both edit an article on God by agreeing to describe what philosophers and theologians have written. NPOV was not neutral about everything. It privileged published sources. But this particular partiality allowed a commons to function.

In April 2002, during a dispute with an anonymous contributor who kept pressing his own views, Wales put the working standard plainly: Wikipedia should take no position "for" or "against" contested views, but should "present those views in such a way that both proponents and detractors can mutually agree."\autocite{WalesNPOVDisquieting2002}

At its core, it required that articles represent all significant viewpoints fairly, without asserting that any particular viewpoint was correct.\autocite{WikipediaNPOV} Wikipedia would describe controversies rather than resolve them. If reliable sources disagreed, the article would present the disagreement. If one view was dominant in published sources, the article would reflect that dominance while still noting alternatives.

\section*{Wikipedia Enters}

Nupedia, the expert-written project whose slow failure \chapref{ch:expert-failure} recounted, had launched in 2000 under a homegrown license of its own, the Nupedia Open Content License.\autocite{en:Nupedia} In January 2001, after email exchanges withcontacting Richard Stallman himself, Jimmyabout licensing, Wales committed Nupedia,his expert-written encyclopedia project, and then Wikipedia, to the GFDL.\autocite{Sanger2005Memoir}GFDL.\autocite{SlashdotGNUpediaStarting2001,WalesSlashdotNupedia2001}

This was not a casual choice. The GNU Project noted: "Just as we were starting a project, GNUpedia, to develop a free encyclopedia, the Nupedia encyclopedia project adopted the GNU Free Documentation License and thus became a free commercial project [sic]. So we decided to merge GNUpedia project into Nupedia."\autocite{GNUFreeEncyclopedia}\footnote{The phrase "free commercial project" appears verbatim on the GNU Project's page; in context it almost certainly means a free-content project.}project. The "merge" was likewise Stallman's framing. No formal merger occurred: Wales already owned the gnupedia.org domain, and Stallman's project, renamed GNE, was shelved once Nupedia adopted the GFDL.\autocite{en:GNE}}

Wikipedia entered Stallman's ideological framework. The project accepted his legal architecture, his philosophical commitments, and his assumptions about how protection would work.

Contributors who edited Wikipedia understood the deal: their work would remain free forever. Anyone could use it, modify it, share it, but derivatives would have to preserve the same freedoms. Commercial publishers couldn't lock up volunteer labor. The copyleft mechanism would ensure that what was shared stayed shared.\autocite{Reagle2010}

Sanger described the bargain in plain terms. Posting text to Wikipedia meant surrendering exclusive control of it. ``I no longer own the rights to the text,'' Larry Sanger wrote on the project's mailing list in June 2001. ``It is now released under the GNU FDL.'' He admitted a private unease the license did not share, that he would be ``a bit miffed if someone started making a bunch of money from it.''\autocite{SangerCopyrightForking2001}

This was the promise.

Stallman had stood before the dead PDP-10, watching a community's destruction. His response was to build legal defenses against future enclosures. Those defenses worked for decades, protecting free software and free documentation from commercial appropriation. The institution that inherited them would be tested by enclosures Stallman never imagined.

In 2001, Wikipedia made a promise to its contributors: your work will remain free forever. That promise wascontributors, encoded in the GFDL, backed by Stallman's legal architecture, inherited from the grief of a programmer who watched his community die. The site's earliest copyright statement said its content "is free and will remain so forever."\autocite{NostalgiaWikipediaCopyright2001}

Tens of thousands of editors would contribute under that promise. Millions of articles would be created. The largest reference work ever assembled would emerge from collective volunteer labor, protected by copyleft.\autocite{WikipediaSizeOf}

When Wikipedia launched, there was no real alternative. Creative Commons would not release its first licenses until December 2002, nearly two years later.\autocite{WikipediaCCLicense} For a project committed to copyleft, where every reuse must carry forward the same freedoms, the GFDL was the only door.\autocite{WMFLicensingQA2009}

JimmyAs \chapref{ch:copyleft-promise} recounted, Wales had corresponded with Richard Stallmantraded Nupedia's homegrown Open Content License for the GFDL inearly January 2001 about licensing.\autocite{Sanger2005Memoir}2001, days after contacting Richard Stallman.\autocite{en:Nupedia,WalesSlashdotNupedia2001} Wales wasnot a free software ideologue. He was an entrepreneur whose web portal, Bomis, ran on advertising revenue.\autocite{WikipediaBomis} But he understood that a free encyclopedia needed a free license, and Stallman was the expert.

As one anonymous Slashdot commenter would later quip, recalling those early days: ``Having Richard Stallman brow beat Jimmy Wales certainly didn't hurt either.''\autocite{slashdot:wikipedia-gfdl-to-cc-2009} It was hearsay, offered eight years after the fact, and even the commenter doubted it changed much.

TheWas the choice was pragmatic, not ideological.conviction or convenience? Wales came to the project having read Eric Raymond's "The Cathedral and the Bazaar," urged it on Sanger, and the two advanced their encyclopedia as an open-source-inspired, "open content" undertaking; Nupedia's own home page called it "the open content encyclopedia" from the spring of 2000.\autocite{SangerInterview2026,NupediaOpenContentHome2000}

Sanger has described his own commitment as conviction, and his advocacy at the time reads that way.\autocite{SangerInterview2026,SangerCopyrightForking2001} Wales's motive is harder to fix: he may have prized open source as a philosophy, as a way to make money, or as both at once; Bomis had shown Waleshim that advertising could fund free content, and the GFDL's copyleft requirements were a constraint he accepted rather than a creed he embraced. Therecord does not settle it. Either way, the tension between free-content ideology and commercial logic was built into the project from the start, and it outlasted any one person.

When Wikipedians debated the license's fine print in the autumn of 2001, Simon Kissane opened the thread by arguing that precision was beside the point. ``Since the FDL is enforced through means of social pressure rather than legal proceeding,'' he wrote, ``its spirit is more important than its letter.''\autocite{KissaneLicensingBother2001} Wales agreed. He confessed the details had never much concerned him: ``We just say `It is released under the GNU FDL' and I didn't worry too much about the exact details.'' He kept the license for a practical reason, ``the instant `free' credibility of the GNU FDL license.''\autocite{WalesLicensingBother2001}

\section*{GFDL Requirements}

Michael Snow, the Foundation's board chair, stressed continuity in announcing the result: "The volunteers who work on Wikimedia projects have very strongly supported making their contributions available under the Creative Commons Attribution/Share-Alike License (CC-BY-SA) in addition to the GNU Free Documentation License (GFDL)."\autocite{WMFPressDualLicense2009} Erik Moeller, then deputy director, called it "a big day for free culture."\autocite{MoellerFoundationL2009}

Critics countered that the vote bound people who never cast one, and that consenting to later versions of GFDL was not the same as consenting to a different license family. The objection predated the ballot: in October 2008 the Norwegian contributor John Erling Blad had told the foundation mailing list, ``you can't relicense content you do not own.''\autocite{BladLicenseSituation2008} Chris Frey had already named the stakes in his letter to the FSF: a "breach of trust."\autocite{fsf:2008-12-fdl-open-letter}

On the technology news site Slashdot, others saw troubling precedent. Anyone using an FSF license with the "or later" clause had just learned that their work could be relicensed without their participation.\autocite{Slashdot2009Migration} The turnout gap, critics argued, made the result morally thin even if procedurally valid.\autocite{wipo:wikimedia-licensing-policy-change-a-conundrum-37077}

The absence of a legal challenge did not settle the question.

\section*{Who Enforces?}

The 2009 migration to CC-BY-SA solved the friction problem.\autocite{wipo:wikimedia-licensing-policy-change-a-conundrum-37077} It did not address the enforcement problem at all.\autocite{meta:GFDL_and_CC-BY-SA_enforcement} Inside the project, the missing enforcer was already a joke. Copying a template between wikis in 2005, the editor Rich Holton asked whether he should ``sleep with my boots on in case the GFDL police come beating on my door.''\autocite{HoltonGFDLPolice2005} There were none.

Consider who might enforce Wikipedia's license. The Foundation cannot sue on the license that matters most. Copyright in each article stays with the editor who wrote it, and the Foundation holds no assignment, only the same public CC-BY-SA license everyone else receives.\autocite{creativecommons:legalcode_en,WMFTermsOfUse2024} Under United States law a bare non-exclusive licensee has no standing to sue for infringement, so the Foundation cannot bring a copyright claim over reused article text. The editors can. No one else can. Meta-Wiki says so plainly: "The Wikimedia Foundation owns almost none of the content on Wikimedia sites---it is owned by the individual creators."\autocite{meta:GFDL_and_CC-BY-SA_enforcement}

But these six pages created the legal vessel for what became the world's largest encyclopedia.\autocite{en:History_of_Wikipedia} Every article Wikipedia's contributors wrote, every edit and image and reference, fell under the stewardship of the entity born that day. The Wikimedia Foundation became the institutional home of free knowledge.\autocite{foundation:2003}

The announcement to the community was brief. Wales explained that he was transferring the project's domain names and the copyrights Bomis held to the new nonprofit. The contributors who wrote the articles kept their own copyrights, licensing their work to anyone under the project's free license. The servers, for the moment, stayed with Bomis, with donation to follow once the tax implications were clear. The Foundation was meant to be an "independent charitable entity," separate from Bomis.\autocite{foundation:2003}Bomis.\autocite{foundation:2003,WalesAnnouncingFoundation2003} The community had worried about commercial pressures, and the Foundation was supposed to protect against them.\autocite{walesWikipediaLMar2002}

Told this way, the founding is a story of generosity: a founder building an institution to serve the commons.

\section*{Asset Transfer}

Wales's decision to createIn creating the Foundation was genuine charitable giving.Foundation, Wales gave up more than he had to.

He transferred the domains (wikipedia.org, wikipedia.com) and the copyrights Bomis held at the founding. The trademarks and server hardware followed.\autocite{foundation:2003} He kept nothing for himself. A less generous founder could have structured things differently, keeping the trademarks, licensing them back, holding control through ownership rather than governance.

The 2003 Foundation was tiny: about \$57,000 in assets, no paid staff, volunteers answering emails.\autocite{en:Wikipedia:Fundraising_statistics} It grew into something more than a thousand times larger. That growth was not corruption. Governing a top-ten global website demanded infrastructure, legal defense, technical work, and community coordination.\autocite{datareportal:digital-2025-exploring-trends-in-wikipedia-traffic} Volunteers alone could not scale to meet the work.

But the architecture set in 2003 shaped everything after it. The clearest sign came in 2008, when the board formalized a Founder's Seat that no one but Jimmy Wales can hold. Not even Larry Sanger, who designed Wikipedia's core policies and served as its first editor-in-chief,chief organizer, qualifies.\autocite{meta:Founder_27s_seat,Sanger2005Memoir} The bylaws encoded what incorporation and funding had already decided: governance followed ownership, not contribution.

The seat gave institutional form to a quarrel the two men never resolved. As \chapref{ch:sangers-departure} recounts, Sanger had claimed co-founder standing since he lefthis paid role ended in March 2002, pointing to the project's name andname, its Neutral Point of View policy.\autocite{SangerNoLongerParticipating}core policies, and the January 2002 press release that had named both men as founders.\autocite{SangerNoLongerParticipating,WikipediaPressReleaseJan2002} From 2006, Wales tiedpublicly disputed the title instead to ownership oftitle, calling the assets heco-founder label ``preposterous.'' In his later telling, Sanger had funded and donated.\autocite{lih2009wikipedia}not founded anything: he was an employee, working under Wales on a project Bomis funded.\autocite{MeheganBostonGlobe2006,lih2009wikipedia} What the Founder's Seat added was an institutional consequence: a chair defined so that the man who wrote Wikipedia's founding rules could never occupy it.

The person who designed the rules was shut out of administering them. It was the early shape of a pattern: an institution created for the community but not by the community. Wikipedia's value came from collective labor, not just Wales's money or Sanger's policies but a quarter million contributors adding knowledge article by article. The institution governing that work was not built by the collective. Its structure reflected legal ownership, and the fight over a title was,reads, at bottom, as a fight over who controlled the result.

\section*{Servers Cost Money}

The stated reasons were practical: a deeper talent pool, proximity to partners, cheaper international travel.\autocite{signpost2007relocation} The unstated aspiration was evident. Wikipedia was joining the tech industry.

San Francisco was not just expensive and ambitious. It was also among the most politically homogeneous major cities in America, a Democratic bastion where professionalAmerica: the year the Foundation arrived, its county voted 84 percent for Barack Obama, and the standard academic ranking of big-city politics would place it first among the fifty-one largest US cities for liberalism.\autocite{CASOSVote2008,TausanovitchWarshaw2014} Professional networks there sorted by ideology as surely as by skill. Hiringskill, and hiring from thesethose networks would shape who staffed the institution.

What gets left behind in a move is not just geography. It is identity. The scrappy volunteer project that ran on donated servers and volunteer labor was becoming something else: an institution with office space, employee benefits, and strategic planning documents.\autocite{lih2009wikipedia}

This was not a software dispute but a governance one. The Foundation had built a mechanism to override consensus by force. A protest letter drew 824 signatures.\autocite{metaletter_superprotect} A vote went 664 to 103 to remove the permission.\autocite{metarfc_superprotect} Superprotect stayed in place for fifteen months. The Foundation removed it on November 5, 2015,\autocite{signpost2015nov} after Executive Director Lila Tretikov acknowledged that it had ``set up a precedent of mistrust, and this is something it was really important for us to remove.''\autocite{mediawiki_wmfprocess_2015}

The third crisis turned on secrecy. In January 2016, details emerged of a Foundation project called Knowledge Engine, billed as ``the internet's first transparent search engine'' and funded by a \$250,000 Knight Foundation grant.\autocite{en:Special_report_2} The community had not been consulted. The Board had not been fully informed.\autocite{meta:James_Heilman_removal_FAQ} Trustee James Heilman was removed from the Board.\autocite{foundation:Resolution:James_Heilman_Removal} He maintained the removal was punishment for pushinghaving ``pushed for greater transparency aboutboth within the project,\autocite{lists:TT4LXY56NGGMLUORCCQBEVIPZ7RCVUK2}WMF and with our communities,''\autocite{lists:TT4LXY56NGGMLUORCCQBEVIPZ7RCVUK2,HeilmanBoardChanges2016} a connection Walesangrily disputed.\autocite{signpost2016feb10}

Staff departed in waves. Luis Villa, Anna Koval, and Siko Bouterse left within days of one another, Bouterse resigning with a pointed public statement. Tretikov resigned on February 25, 2016.\autocite{en:Special_report_2} An organization built on openness had developed its central strategy in secret.

Twenty-one administrators resigned in protest. The discussion ran to 470,000 words.\autocite{slate_fram_2019} ``I am not willing to serve Jan and the T\&S team... under undisclosed new rules and under threat of unappealable sanctions,'' one resigning administrator wrote.\autocite{wp_fram_community_response} By late July 2019, the active administrator count had dropped below 500 for the first time.\autocite{signpost_admin_decline_2019} In September, ArbCom vacated the ban.\autocite{signpost_arbcom_unbans_2019}

Four crises, three executive tenures, one pattern. VisualEditor unfolded under Sue Gardner, Superprotect and the Knowledge Engine under Lila Tretikov, the Fram ban under Katherine Maher.\autocite{wikipedia_wmf_org} None of the fourfour, on this book's reading, was a personal failing. Each director met the same structural bind: the staff held the resources and the legal authority, the community held only its voice. The tension persisted no matter who led. This was mission drift in miniature: an organization built to serve editors had grown priorities of its own, and when those priorities collided with the community's, the staff held the power to win.

\section*{Diverging Trajectories}

The Croatian press first raised the alarm in February 2012.\autocite{hrwiki_self_history} In September 2013 the story went national: a front-page investigation in \textit{Jutarnji list} under the headline ``Right-wingers took over the editing of Croatian Wikipedia,'' the minister's warning three days later, a protest Facebook page that drew thousands.\autocite{penic_jutarnji_ndh_2013,tomicic_novilist_2013} Jimmy Wales told a Croatian daily that year that the Foundation was aware of the problems and looking into them.\autocite{wmf_croatian_assessment_report_pdf_2021} A formal request for comment opened on Meta-Wiki that same month, ran for roughly six months, and closed without result. The Foundation's own report later supplied the epitaph: ``There was no Foundation process in place backstopping the struggling community effort.''\autocite{wmf_croatian_assessment_report_pdf_2021,meta_rfc_2013_croatian}

In October 2013 a Croatian editor, Dalibor Bosits, brought the case to the movement's own mailing list.\autocite{wikimedialist_croatian_2013} He described administrators who had ``been operating as a cabal of sorts'' and wrote that the community was ``unable to resolve the problems without your assistance.'' The first reply did not dispute the facts. It asked whether the Foundation ``can and should intervene in project disputes when one side makes a request.''\autocite{wikimedialist_croatian_reply_2013}

The alarms kept coming. In 2018, Balkan Insight published ``How Croatian Wikipedia Made a Concentration Camp Disappear,'' and Novosti published its diff-level investigation under the title ``Endehapedia,'' after the Ustaša state's initials.\autocite{milekic_balkaninsight_jasenovac_2018,krnic_novosti_endehapedia_2018} A prominent Croatian writer concluded the project could be called, ``without being unjust or exaggerating,'' an Ustasha Wikipedia.\autocite{signpost_curious_case_2019}

Copyleft had always assumed that derivatives would be visible. AI training hid them instead. Model weights were not articles, and training looked nothing like copying. But the copies were there, embedded in the parameters and recoverable by anyone who prompted for them. The copying was real. What failed was the accounting: an industry that preferred not to look, and a steward that never forced it to.

The commons itself did not notice the taking. The training table sat in public from May 2020, but across the movement's own mailing lists, nearly half a million messages through early 2021, GPT-3 drew two passing mentions. The first dedicated thread came on December 10, 2022, ten days after ChatGPT launched, opening with a question: ``Is this Ai software using info from Wikipedia directly or indirecly?''\autocite{WennerstenChatGPT2022}

The first reply, within the hour, named the stakes: ``I would argue it's violating the copyright,'' wrote the editor known as The Cunctator, linking the GPT-3 paper, ``but I'm aware the foundation isn't very interested in defending it.''\autocite{CunctatorChatGPT2022} Sixteen more replies followed, curious and technical.

Erik Moeller, the Foundation's former deputy director, tested an early chatbot search engine and reported it ``basically rewriting/summarizing the Wikipedia article''; his objection was accuracy, judging it ``irresponsible to market it as part of a search engine in its current form.''\autocite{MoellerChatGPT2022} The extraction had been public for two and a half years before the community asked whether it had happened, and the first answer named both the violation and the odds of anyone acting on it.

The GPT-3 paper appeared on May 28, 2020. Five years later, Wikipedia topped ChatGPT's citation lists, accounting for nearly half of its most-cited sources.\autocite{tryprofound:ai-platform-citation-patterns} The editors, the arbitrators, the tireless fixers of vandalism saw none of the benefits. No consent, no consultation, no visible credit.

From inside the licenses, the gap was a conflict of interest. CC-BY-SA required attribution and share-alike,\autocite{creativecommons:legalcode_en} and the Foundation had already admitted, in March 2024, that many AI companies probably honored neither.\autocite{jillianbommarito:wikimedia-says-no-llm-training} Enforcing the license meant suing the violators. Of the likely violators, OpenAI was the one company whose suit would cost the Foundation no revenue, because OpenAI paid none.\autocite{enterprise:wikipedia-25-enterprise-partners} The companies that did pay might be breaking the same terms. Suing them would end the revenue. So the Foundation sued no one.

That was the equilibrium the structure produced. Move against OpenAI, and the paying customers ask why they were singled out. Move against the payers, and the revenue collapses. Universal non-enforcement was the only stable outcome, dressed as pragmatic adaptation.outcome.

An adversarial organization would have built enforcement before taking corporate money. The Foundation never did. Its leadership moved in the same circles as the executives it negotiated with, sharing conferences andconferences, a vocabulary of access and scale.scale, and, as the donation records ahead show, the same politics.\autocite{FECScheduleA2026} Enterprise formalized a relationship the leadership already consideredtreated as natural. FinancialThe dependency that followed grew from culturalthat affinity, not the reverse.reverse; the chapters ahead lay out the evidence for that argument.

\section*{Who Paid, Who Used}

The absences were more telling than the list. Google had almost certainly drawn more from Wikipedia than any AI company, feeding it through two decades of search and now into Gemini, yet Google had at least become a paying customer. The heaviest model-maker, OpenAI, had not, and neither had Anthropic. Both used Wikipedia freely and paid nothing.\autocite{enterprise:wikipedia-25-enterprise-partners} Neither did xAI, whose Grokipedia had forked Wikipedia, copying it wholesale, in late October 2025.\autocite{PBS2025Grokipedia}

The more revealing contrast was between the two absentees. OpenAI ignored the license's terms entirely and drew no rebuke beyond Wales's occasional comment.\autocite{enterprise:wikipedia-25-enterprise-partners,jillianbommarito:wikimedia-says-no-llm-training} xAI, whose fork at least credited Wikipedia and carried its license forward, was about to become the Foundation's loudest grievance.\autocite{PBS2025Grokipedia,NBC2025NeoNazi} The full irony of that pairing belongs to Grokipedia's story, two chapters on. But the direction was already visible:visible, and it raised a question the coming chapters take up: whether the criticism tracked ideology more than compliance.

The structure explained the silence. Once the Foundation drew profit from AI-company revenue, legal action against AI companies became implausible, whether or not they were the ones paying.\autocite{diff:wikimedia-enterprise-financial-report-fiscal-year-2024-2025} The arguments for enforcement already existed. The record held the memorization research showing that models reproduce their training text word for word, the market-harm findings, and the Foundation's own admission that few AI companies honored the license.\autocite{Carlini2021Extracting,Carlini2023Quantifying,diff:new-user-trends-on-wikipedia,jillianbommarito:wikimedia-says-no-llm-training} What was missing was the will, and the structure had removed it.

The Wikimedia Foundation had no equivalent enforcement constituency. Political homogeneity had eliminated the internal voices that might have argued for adversarial enforcement. Nearly every donating employee gave to the same party, and the leadership pipeline drew from the industry's own pools. Partnership felt natural because, culturally, it was. Cultural affinity preceded financial entanglement. By the time Enterprise revenue created dependence, the Foundation had already decided these companies were allies.

That homogeneity had a cost beyond licensing.A culture that generated no internal argument for enforcement also tolerated little dissent of any kind. The community demonstrated as much in June 2026, when it blocked the co-founder indefinitely.\autocite{Sanger404Media2026} Larry Sanger had spent years arguing that Wikipedia's culture was politically lopsided\autocite{larrysanger:wikipedia-is-badly-biased} and needed an organized counterweight.\autocite{thefp:i-founded-wikipedia-heres-how-to-fix-it} WhenIn June 2026, when he tried to build one, the machinerycommunity blocked him indefinitely, citing off-wiki canvassing, a contribution record thin on articles, and battleground conduct.\autocite{Sanger404Media2026,SangerANI2026} The procedural case was not frivolous; Sanger himself conceded that one recruiting post ``contains indefensible rhetoric, and I do regret it,'' even as he disputed the fairness of the process that judged him.\autocite{SangerANI2026} Machinery that never moved against the extractors moved brisklywithin days againsthim.\autocite{Sanger404Media2026} A community that would not abide a loyal opposition on politics was not going to growcritic's canvassing. This book reads the two monocultures, the oneon licensing. The monoculture that disarmed enforcement and the monocultureone that expelledits co-founder were the same.co-founder, as the same one.

\section*{The OpenAI Paradox}

\section*{The Freedom to Fork}

One voice complicated the response to Grokipedia. Larry Sanger, Wikipedia's estranged co-founder, had criticized the encyclopedia for years, arguingyears; in a May 2020 essay, "Wikipedia Is Badly Biased," he argued it had tilted toward mainstream establishment views.\autocite{ViceSanger2015,larrysanger:why-wikipedia-must-jettison-its-anti-elitism}abandoned its own neutrality policy.\autocite{larrysanger:wikipedia-is-badly-biased} When Grokipedia appeared, he was cautiously supportive: "The more encyclopedias in the world there are, the happier I am."\autocite{SAN2025Sanger}

His first read was measured: "My initial impression, looking at my own article and poking around here and there, is that Grokipedia is very OK. The jury's still out as to whether it's actually better than Wikipedia. But at this point I would have to say 'maybe!'"\autocite{SangerX2025Initial}

Sanger was not blind to the failures. He called the AI's confident invention of false information "bullshittery."\autocite{PCMagSanger2025} He graded his own Grokipedia entry a "C," "not at all a failure" but with "many bad points." He noted bias running in more than one direction, including to the right.\autocite{SangerGrokipedia2025} He judged Grokipedia as he would any encyclopedia, by content quality rather than political valence.

His view rested on a principle the copyleft movement had always held: the freedom to fork is fundamental. The Spanish Wikipedia community had forked in 2002 over commercialization fears.\autocite{cpov2011spanishfork} Sanger's own Citizendium had forked in 2006 over expertise and governance.\autocite{lih2009wikipedia} Citizendium had failed for want of contributors, though one fork in one decade proved little on its own, as \chapref{ch:sangers-departure} cautioned. Grokipedia was the retest: machine generation had drained the moat of labor that had protected the incumbent. The license permitted these forks by design.\autocite{creativecommons:legalcode_en} When governance failed, the remedy was to build something else.

The open question was whether Grokipedia's content failures disqualified it as a legitimate fork. Sanger thought the jury was still out. The neo-Nazi citations and Holocaust minimization suggested otherwise. But those were editorial failures, the kind better curation could correct. They were not structural violations of what copyleft protected.

Within months, that principle cut against Sanger himself. In June 2026 Wikipedia's community blocked his account indefinitely, after a dispute over his campaign for a reform project and his off-site recruiting for it.\autocite{Sanger404Media2026} He had argued for decadesmade that failed governance justified building an alternative.\autocite{SangerNoLongerParticipating}case in practice before, leaving in 2003 and founding Citizendium in 2006.\autocite{SangerNoLongerParticipating,lih2009wikipedia} The encyclopedia he named appliedcharges were conduct rules, not his own logic toforking philosophy turned against him, leavingand he disputed the freedomprocess itself: ``no due process, no prosecutor, no dispassionate judge, no jury,'' he wrote the morning of the block.\autocite{SangerX2026Block} What remained to fork ashim was theonly freedom he had left.the license guaranteed: the freedom to fork.

\section*{The Loop Begins}

\begin{table}[htbp]

\centering

\begin{tabularx}{\textwidth}{Xrl}\begin{tabular}{lrl}

\toprule

Metric & Value & Source \\

\midrule

New webpages containing AI-generatedpages with AI content(April 2025) & 74.2\% & Ahrefs\autocite{AhrefsAIContent2025} \\

AI-writtenAI pages in top-20top Google results & 11\% to 20\% & Originality.ai\autocite{OriginalityAIGoogle2025} \\

AI ``news'' sites(since 2023) & 49 to 1,271 & NewsGuard\autocite{NewsGuardAITracking2025} \\

Projected AI-generated online contentAI share by 2026 & up to 90\% & Europol\autocite{EuropolDeepfakes2022} \\

Hallucination rate, one frontier model (benchmark)Frontier-model hallucination rate & 79\% & OpenAI\autocite{OpenAIo3o4SystemCard2025} \\

\bottomrule

\end{tabularx}\end{tabular}

\caption{AI-generated content contamination metrics}

\label{tab:ai-contamination}

\end{table}

The rows measure different things. Ahrefs sampled webpages newly published in April 2025 and found AI-generated text in three of every four. Originality.ai measured how much of Google's top twenty results was AI-written. NewsGuard's count of AI-run ``news'' sites grew from 49 to 1,271 in under three years.

Two of thosethe numbers were estimates rather than measurements: the 90 percent was a forecast,Europol's forecast for online content by 2026, and the 79 percent came from one frontier model on a hard factual benchmark. The direction they pointed to was not in dispute. By 2025, most new web content carried the mark of AI, and Wikipedia's seven million human-curated English articles were an island in an ocean of that content.\autocite{WikipediaSizeOf} That ocean was feeding back into the AI systems that had created it.

\section*{What the Collapse Research Missed}

The degradation loop was not just a technical failure. It was the logical endpoint of non-enforcement, the consequence of a compromised steward. When the institution meant to protect the shared resource became financially dependent on those harvesting it, the resource lost its only defense. The copyleft Stallman built as armor sat unworn.

Enforcement was expensive, but the deeper failure was institutional will: no one inside the institution saw a reason to put itthe armor on. Political homogeneityhomogeneity, measurable in the staff's own donation records,\autocite{FECScheduleA2026} had bred out adversarial instincts. ToFor an institution staffed by people who sawdrawn from the executivessame milieu as peers,the executives across the table, ``partnership'' was a default, not a decision. The defense failed inside the institution meant to provide it.

This was mechanized loss, damage that perpetuated itself without further human intervention. Each training cycle absorbed the errors of the last. Model weights could not be "cleaned."\autocite{Shumailov2024ModelCollapse} Human editors could correct Wikipedia. They could not correct the weights of systems they had no access to. The harm compounded on its own. The off switch did not exist.

From this view, Wikipedia's adaptation to AI was working. The Pragmatists were keeping the lights on.

The camp's clearest statement came from Samuel Klein, a former trustee, in July 2026. An AI system "knocking on your library window a billion times a week," he wrote to the movement's list, "is not a burden, it's a relationship." Wikimedians owed "a responsibility toward AI systems that rely on Wikipedia."\autocite{KleinWikimediaL2026} His program was the opposite of a sellout: open weights, open data, public AI coalitions. But the responsibility he named ran toward the extractors, and his long list of references did not mention the license once. As of this writing, the post has drawn no reply on the list.

\section*{The Copyleft Purists}

One more barrier ran beneath the rest: the monoculture. The employee donations ran more than 99 percent to one party,\autocite{FECScheduleA2026} the professional networks overlapped with the AI executives' own, and the vocabulary matched. Nobody pushed for adversarial enforcement because nobody inside was built to.

The same logic shaped enforcement inside the community. In June 2026, as \chapref{ch:sangers-departure} recounts, editors invoked WP:BATTLEGROUND, the rule against treating Wikipedia as a contest to be won,\autocite{WikipediaNotBattleground2026} to block co-founder Larry Sanger indefinitely for organizing an openly declared faction, WikiProject Intellectual Diversity, and recruiting for it off-wiki.\autocite{Sanger404Media2026,SangerANI2026} The chargeposts cited in the June 2026 noticeboard discussion were real, and Sanger's own accounts of the campaign were public; the dispute was not baseless: steering an outside audience into internal votes isover what they meant.\autocite{WikipediaCanvassing2026}

The community's case had substance: he had made fewer than fifty article edits since 2003, and Sanger himself conceded that his ``march right back'' post contained ``indefensible rhetoric,'' a real violation.\autocite{WikipediaCanvassing2026}regret he put on the record in his own defense.\autocite{SangerANI2026}

But the asymmetry was procedural. WP:BATTLEGROUND could only be invoked against a side that had declared itself a faction. The incumbent alignment never had. It needed no WikiProject and published no manifesto. It worked through the ordinary machinery of consensus, in a culture whose measurable political donations ran almost entirely one way, and so it never registered as a faction the rule could reach.\autocite{FECScheduleA2026} The community's own enforcement tools could see only the side organized enough to name itself.

This time, though, the exit was closed. Wikipedia's dominance was far more complete than in 2002, its brand synonymous with ``encyclopedia.'' More than 65 million articles across languages could not be easily rebuilt.\autocite{wikimediafoundation:wikipedia-celebrates-25years} Any fork would face the same AI extraction, because scraping did not depend on who hosted the content. And organizing a quarter million editors was a different task from organizing a few dozen Spanish contributors.\autocite{cpov2011spanishfork}

Forking solved nothing. Staying meant accepting capture. Sanger had tried both,both: the fork that never scaledscaled, and the reform that got him blocked, andeffort the community ended with neither door open.\autocite{CitizendiumWikipedia,Sanger404Media2026}his June 2026 block.\autocite{CitizendiumWikipedia,Sanger404Media2026}

\section*{The Civil War Already Here}

The parallel between Symbolics and the AI companies is close in the ways that matter here. In 1982 the commons was Lisp Machine code.\autocite{Williams2002Freedom} In 2024 it was the English Wikipedia's 6.7 million articles.\autocite{WikipediaSize2026} The contributors were MIT hackers, then hundreds of thousands of editors.\autocite{WikipediaAbout2026} The promise was a gentleman's agreement to share improvements, then the ShareAlike clause of CC-BY-SA.\autocite{creativecommons:legalcode_en} The betrayal was the same in both: take the shared work, lock the improvements away, give nothing back.

Only the response differed. What changed was disposition, not circumstance. Stallman treated corporations as adversaries by default.\autocite{Williams2002Freedom} He assumed betrayal and built his defenses before it came. The Wikimedia Foundation treatedassumed partnership and signed deals with the companies extracting its work as peers.\autocite{techcrunch:wikimedia-foundation-announces-new-ai-partnerships-with-amaz}work.\autocite{techcrunch:wikimedia-foundation-announces-new-ai-partnerships-with-amaz} When the AI firms arrived to extract, they found colleagues, not adversaries.no adversarial posture met them.

The two responses are worth setting side by side. When Symbolics betrayed the community, Stallman cut their link, declared war, and spent two years keeping the lab's version alive without them. Then he built the legal machinery to stop it happening again.\autocite{Williams2002Freedom}

The wall Stallman built still stood. The legal mechanism remained available. Enforcement only needed an enforcer with resources, will, and independence. The Foundation had the resources, more than \$180 million in annual revenue.\autocite{en:Wikipedia:Fundraising_statistics} It lacked the will. And it was not independent. Its leadership was cut from the world of the firms taking the work.

The wall was never breached. The evidenceevidence, from the donation records to the partnership deals,\autocite{FECScheduleA2026} points less to a simple purchase than to prior alignment. The guards had already chosen a side.

The question the commons leaves behind is not whether the license could have protected it. The question is why the people charged with defending it chose not to try.

\section*{The Unkept Promise}

If enforcement was possible, and the evidence says it was, then non-enforcement was a choice. Not passive drift. Not institutional inertia. A decisionchoice made at specific moments by specific people who judged thatdatable moments: when Enterprise revenue mattered more than licensing integrity.launched without an enforcement arm in 2021, when the March 2024 admission changed no contract, when each renewal was signed anyway.\autocite{openfuture:wikimedia-enterprise,jillianbommarito:wikimedia-says-no-llm-training} What the people in those rooms weighed, no outsider can see. What they chose is on the record.

That decision had a cost, and the people who paid it never agreed to it. Volunteers built Wikipedia under the promise of CC-BY-SA, and a quarter million of them still maintain it.\autocite{WikipediaAbout2026} Their work would stay free. Derivative works would be shared alike. Attribution would follow their words wherever those words traveled.\autocite{creativecommons:legalcode_en}

No one asked those volunteers when the Foundation signed Enterprise deals. No one asked them when the Foundation admitted, in 2024, that most of its customers were not complying with the license.\autocite{jillianbommarito:wikimedia-says-no-llm-training} No one represented them when revenue won out over the license they had been promised.

The case for enforcement was there to be made. The people who could have made it answered tonow had customersnow, not to answer to, not the contributors whose promise went unkept.

In 2024 and 2025, that bargain visibly shattered. Volunteers saw their words reproduced in locked proprietary systems. They saw users asking ChatGPT instead of visiting Wikipedia.\autocite{datareportal:digital-2025-exploring-trends-in-wikipedia-traffic} When the Foundation signed its Enterprise data-feed deals,\autocite{techcrunch:wikimedia-foundation-announces-new-ai-partnerships-with-amaz} volunteers discovered they had not been consulted. The institution they had built had grown dependent on the companies extracting their labor.

The pattern was not new. In 2020, the Foundation pushed to rebrand the entire movement around the Wikipedia name. The editor David Gerard dismissed its \enquote{community consultation} as \enquote{window dressing for a decision that's long been made.}\autocite{Gerard2020Branding} The Board paused the project that September.\autocite{Sefidari2020BrandPause}

When the Foundation tested AI summaries in June 2025, volunteers killed the experiment within a day, and they rejected Wales's proposal to use ChatGPT for article reviews the month after.\autocite{404media:jimmy-wales-wikipedia-ai-chatgpt} They were not opposed to AI in principle. They were opposed to being extracted without consent.

\end{table}

These were the documented starting conditions, and the community that read them had already split four ways over what they meant. The pragmatistsPragmatists saw inevitability: AI was here, so the Foundation should be paid for access. The copyleft puristsCopyleft Purists saw the founding promise being broken. The skepticsSkeptics saw an existential threat in AI itself. The volunteer advocatesVolunteer Advocates saw their labor turned to someone else's profit without consent. Each camp read the same numbers and saw a different path forward. None commanded consensus.

The question was not whether Wikipedia would survive. The encyclopedia would persist in some form. Whether a commons could survive capture was the harder question.

The scenarios are not alternatives but simultaneous trajectories. Their relative strength will decide what comes. By mid-2026 the third had its first real test on the docket: the Third Circuit heard the first federal appellate argument on fair use in AI training that June, in the ROSS appeal, with the decision pending.\autocite{ThirdCircuitRossDocket2026}

The futures were still arguments when the institutionmovement revealed its character. On June 22, 2026, as \chapref{ch:sangers-departure} recounts, the community blocked Larry Sanger, the co-founder who had named the encyclopedia.\autocite{Sanger404Media2026,Sanger2005Memoir} The block had a stated basis.basis, and it came from the volunteer noticeboard, not the Foundation; Wales himself called an indefinite ban ``ludicrous.''\autocite{JimboWalesTalk2026} But the same institution thatmovement's two arms displayed one asymmetry. The community produced a verdict against a critic with no institutional standing in days. The Foundation could not summon the will to enforce its licenseproduce an enforcement letter against companies worth trillions summoned it, within days, against a critic whose only power was his voice. The institution keeps its convictionin years. Conviction for the weak and itsweak, accommodation for the strong. Whichever scenario prevails, that character will shape it.

Elinor Ostrom won the Nobel Prize for showing that communities can govern shared resources without the state or the market.\autocite{NobelOstrom2009} Her eight design principles explained why some commons endured and others collapsed: clear boundaries, rules fit to local conditions, collective decision-making, monitoring, graduated sanctions, conflict resolution, the recognized right to self-organize, and nested layers of governance.\autocite{OstromGoverning1990} Wikipedia met nearly all of them.

You can learn a lot about an institution from the rules it creates and the way it chooses to enforce them, or not.

In March 2024, the Wikimedia Foundation admitted in writing that many of the AI companies using Wikipedia might be violating its license.\autocite{jillianbommarito:wikimedia-says-no-llm-training} That license carried the promise the whole project rested on: what is built from shared work stays shared. The Foundation enforced nothing, and its commercial arm kept collecting from the industry it suspected.\autocite{techcrunch:wikimedia-foundation-announces-new-ai-partnerships-with-amaz} In June 2026, the same institutionmovement expelled the man who had co-founded Wikipedia and given it its name.\autocite{Sanger404Media2026} His offense, in the end,this book will argue, was disagreement.

To understand those two choices, it helps to begin with a copy.

Which brings us back to the man who named it.

On June 22, 2026, an administrator blocked Larry Sanger's Wikipedia account indefinitely.\autocite{SangerANI2026,Sanger404Media2026} Sanger had co-founded the encyclopedia in 2001 and given it its name.\autocite{Sanger2005Memoir} No court ordered the block. No government, no corporation, no hostile takeover. Wikipedia's own community did it, through its own machinery: a noticeboard thread, a tallied ``consensus,''consensus, an administrator enacting it with a single click. ``All my judges,'' Sanger wrote that morning, ``were self-selected and hated me.''\autocite{SangerX2026Block}

The stated charge was off-wiki canvassing.\autocite{SangerANI2026} Sanger had used his following on X and other public platforms to pull outsiders toward an internal Wikipedia vote.\autocite{Sanger404Media2026} Wikipedia's rules forbid recruiting an outside audience into its consensus, and the community enforced them. The breachconduct was real,documented, and the rule is a reasonable one. But strip away the procedure and the substance is plain: Sanger was removed for \emph{disagreement}, forlooks like \emph{disagreement}: years of attacking the project's direction for years and organizing,and, at last, organizing to change it.\autocite{ViceSanger2015}

He had copied nothing, enclosed nothing, broken no one's license. The companies the Foundation served had done all three. They breached the letter and the spirit of the copyleft license that was supposed to keep Wikipedia free, a rule the whole commons rested on, and they remained paying customers.\autocite{jillianbommarito:wikimedia-says-no-llm-training,enterprise:wikipedia-25-enterprise-partners} The dissenter lost his account. The license-breakers kept theirs.

Faced with apowerless critic who disagreed,held no office and no seat, the institutioncommunity found the will to act, decisively, within days. Faced with the most powerful companies on earth breaking the license that defined it, itthe project, the Foundation found only words. ItDifferent bodies, one movement: it had learned to discipline dissent inside its walls and to bill the forces dismantling it from without.

\scenebreak

The second capture followed from the first. Having lost the independence to see the AI industry as anything but partners, the Foundation could not resist it when it came for the commons. Stallman built copyleft to keep the enclosers out; he had not planned for a steward that would decline to enforce it.

The promise began in January 2001, with two founders, an argument about expertise, and a wiki that was never supposed to become the encyclopedia.\autocite{Sanger2005Memoir} One of those founders would spend the next quarter century watching the project driftas its most persistent public critic, arguing from outside for everything he believed it should be.have been. In the summer of 2026, it cast him out.\autocite{Sanger404Media2026}

\noindent The information in this book is provided for educational and informational purposes only, and portions of the research and editing were conducted with the assistance of artificial intelligence tools. While the author has made every effort to ensure accuracy, the author disclaims any liability for errors, omissions, or outcomes arising from the use of information contained herein.\\[0.2em]

\noindent Cover design and typesetting by the author. Printed in the United States of America.\\[0.2em]America. \textbf{Second printing (revised), July 2026.}\\[0.2em]

\noindent \textbf{Book Website:} \texttt{commonscaptured.com}\\

\noindent \textbf{Author Website:}\textbf{Book:} \texttt{commonscaptured.com} \; $\cdot$ \; \textbf{Author:} \texttt{michaelbommarito.com}\\[0.2em]

\noindent ISBN 979-8-951640-00-0 (paperback)\\

% COVER VARIABLES - Amazon KDP Paperback (The Commons Captured)

% ============================================================================

% AUTO-GENERATED by scripts/update_kdp_cover_vars.py on 2026-07-062026-07-19 — do not hand-edit.

% Platform: Amazon KDP | Binding: Paperback

% Paper: white | Pages: 275289 (from main.pdf)

% Spine width: pages * 0.002252 + 0.06 = 0.679300in0.710828in (house formula)

% Layout: [Bleed 0.125] [Back 6.0] [Spine 0.6793]0.7108] [Front 6.0] [Bleed 0.125]

% ============================================================================

\newlength{\CoverSpineWidth}

\setlength{\CoverSpineWidth}{0.679300in}\setlength{\CoverSpineWidth}{0.710828in}

\newlength{\CoverWidth}

\newlength{\CoverHeight}

\setlength{\CoverWidth}{12.929300in}\setlength{\CoverWidth}{12.960828in}

\setlength{\CoverHeight}{9.250in}

\newcommand{\CoverPageCount}{275}\newcommand{\CoverPageCount}{289}

\newcommand{\CoverPaper}{white}