Excavating Early Wikipedia

The encyclopedia discovered its rules by living inside them first

Photo by David Clode on Unsplash

The surviving record of Wikipedia’s first edit, dated 15 January 2001, comes from a reconstruction assembled years later.

For its entire first year the site ran on a Perl program called UseModWiki, which stored each page as a plain text file and discarded old revisions after roughly two weeks. The earliest versions of much of the encyclopedia were therefore gone before anyone thought to preserve them.

When the project moved to new software in early 2002, even the history UseModWiki had kept was left behind at first and was only restored to the database that September. What stayed lost for good was the deepest layer, the revisions the old engine had already thrown away.

The founding act of the largest reference work ever compiled reaches us the way a body does once the flesh is gone. What survives is the hard part, pressed into stone, while the soft tissue that did the living decayed before the record could hold it.

We have one snapshot taken around December 2001, a frozen version editors now call the Nostalgia Wikipedia, along with a scatter of pages whose timestamps were stitched back together when the database changed hands. Everything earlier has to be read the way a paleontologist reads, with the animal inferred from the prints it left in ground that hardened around them.

The reconstruction gets easier as time goes on. Each rewrite of the software kept more, so the record thickens from a few faint impressions in 2001 into a dense, well-dated fossil bed by 2003.

The earliest months are genuinely lost. The middle years are well preserved. Once your reading adjusts to that gradient, the shape in the rock turns out to be stranger than the founding story leads you to expect.

The wager

The ambition was old. A universal encyclopedia, open and shared, had been imagined long before any machine could attempt it.

Paul Otlet began building toward it in Brussels in the 1890s, cataloguing the world’s publications on index cards, work he would later gather into a paper city of knowledge he called the Mundaneum. H. G. Wells proposed a microfilm “World Brain” in the late 1930s.

As scholar Joseph Reagle has observed, every generation reached for whatever medium was newest, and every time the dream outran the tools.

What finally closed the gap was a modest piece of software written, of all things, for programmers.

In 1994 and 1995, a developer named Ward Cunningham wanted an easier way for software engineers to trade ideas about design patterns, so he built the plainest tool that could do the job.

He named it the WikiWikiWeb after the Wiki Wiki Shuttle he had once caught at the Honolulu airport, borrowing the Hawaiian word for quick, and he put it online on 25 March 1995 beside a collection he called the Portland Pattern Repository.

His stated aim was the simplest online database that could possibly work, and two choices folded into that simplicity would shape everything downstream. Anyone could edit any page, with no account and no permission sought. And every page kept a running history, so any change could be undone.

Cunningham added a third commitment, social rather than technical: a standing trust that contributors meant well. The machine assumed that most people, handed an open page, would leave it better than they found it.

Vannevar Bush, in a 1945 essay for The Atlantic, imagined a desk-sized machine he called the memex, built to store documents on microfilm and link them by association so the trails one reader blazed could be handed to the next. Bush had only a description, never a working device, yet his core image, knowledge as a mesh of links rather than a shelf of sealed volumes, kept searching for a body.

It found a partial one in 1987, when Bill Atkinson’s HyperCard arrived on the Macintosh, letting an ordinary person build stacks of linked cards, scriptable and simple enough that schoolchildren took to them.

Cunningham had seen it before the public did, shown a pre-release build by his collaborator Kent Beck while the software still went by the name WildCard, and by his own account it floored him. He used it to make a private stack tracking the engineering ideas at his company, the people who carried them, and the projects they moved through.

When he rebuilt that experience on the web, he swapped the word “pattern” in for “idea,” and People, Projects, and Patterns became the three headings on the front page of the WikiWikiWeb. The wiki’s first table of contents was a HyperCard stack grown up.

What HyperCard could not do turned out to be the whole point. It ran on one machine for one user, so the deepest part of Bush’s vision, that the notes of one mind might travel to another, had nowhere to go. The web was the road.

Cunningham supplied the last missing piece in March 1995, when he lifted a private linked notebook into public space and then took the step no one had quite dared, letting any passing stranger edit it. That single move, opening the notebook to everyone at once, is the moment the wiki became what it is.

That is the inheritance the encyclopedia received. When Wikipedia launched in 2001 it adopted the wiki whole, running at first on an engine descended from Cunningham’s, and the lineage it took on reached back through HyperCard to a magazine essay written before the transistor.

Its real novelty was to aim that old and humble machinery at a target vastly larger than software patterns.

The social rules came down the same line of descent. The wiki culture that grew from Cunningham’s spent the late 1990s working out the etiquette of open collaboration, and MeatballWiki, a site started in 2000 to discuss exactly that, supplied much of the vocabulary that crossed into the encyclopedia almost intact: assume good faith, the principle of first trust, forgive and forget. Soft security sat at the center, inverting the instinct most systems follow.

Rather than guarding the doors with passwords and permissions, you leave them open and let the social fabric absorb the damage afterward. The bet was that helpful hands would outnumber harmful ones, and that whatever slipped through could be reversed.

Reversibility is what kept the wager from being reckless, and it ran on Cunningham’s page history. Because a page kept its recent versions, an edit could be rolled back in seconds, and the project could heal faster than it could be harmed.

Openness was the visible feature. The capacity to undo, working underneath it, was what let openness survive contact with the public.

What it felt like to edit then was stranger than today’s smooth interface suggests. For the first weeks, links were still formed in Cunningham’s CamelCase, capitalized words run together into a single clickable title, so that WorldBrain became a link by typography alone.

The free links in double brackets that editors rely on now were switched on and recommended only in mid-February 2001, a month into the project’s life. Pages looked half-finished because they usually were, and that was the point.

The entrepreneur and the editor

Wikipedia did not appear out of the wiki on its own. It grew off the side of an earlier and more cautious project called Nupedia, an online encyclopedia written by credentialed experts and cleared through a multi-step review, launched in March 2000 and paid for by Bomis, the web company Jimmy Wales ran.

Nupedia was rigorous and painfully slow, managing only a dozen or so finished articles in its first year. After programmer friend Ben Kovitz described Cunningham’s wiki to him in early January 2001, Nupedia’s editor, Larry Sanger, proposed adding one as an open feeder, a rough yard where anyone could draft material that experts might later refine.

The feeder outgrew the parent almost at once. Wikipedia opened on 15 January 2001, and within weeks it was the larger thing.

The two founders brought different gifts. Jimmy Wales, an internet entrepreneur who had traded options before moving online, supplied the money and the servers, and held the longest view of what the project might become.

He stated the goal in the largest terms he could find, the line he would repeat for years about a world in which every person on the planet has free access to the sum of human knowledge. After the founding period he stayed on as the project’s steward and its public face, and he later moved it out of company hands and into a nonprofit built to outlive any owner.

Larry Sanger, a philosopher by training, ran the daily work and shaped the early character. He had written Nupedia’s review rules; when the wiki opened he became its first editor and its only paid editorial hand through the founding year.

He gave the project its name, and his fingerprints are on the earliest policies, from the original guard against bias to the cheerful instruction to ignore the rules. His position lost funding in 2002, as the downturn reached Bomis, and he left the encyclopedia he had named.

Neutrality, the rule that ran backward

One commitment sat at the center from the beginning: the refusal to take sides. The policy now called the neutral point of view did not begin on Wikipedia at all.

It started as a “nonbias” rule that Larry Sanger drafted in 2000 for Nupedia. When the open site launched, the idea reappeared in plainer dress as one of a handful of informal “rules to consider,” under the heading “avoid bias,” with the suggestion that contested claims be attributed to the people who held them.

A fuller statement followed at the end of December 2001, and the name that stuck, neutral point of view, carried a subtlety its critics often missed.

Neutrality, as the early community framed it, never claimed that an objective view from nowhere exists. It was a working procedure for living together. The instruction was to lay a disagreement out and leave it open, recording what people had said about a subject while declining to rule on which of them was right.

That let a believer and a skeptic edit the same paragraph without either having to surrender, because the paragraph described the argument rather than ending it. The project came to treat the policy as non-negotiable for a structural reason. Neutrality was the treaty that let people who agreed on almost nothing keep working in the same room.

Nearly every rule in the story was named after the behavior. Neutrality ran the other way. It was written down as deliberate policy, by one person, at Nupedia, before the open wiki and its swarming conduct existed, and the culture then formed around the rule instead of preceding it.

The principle of the structure is the exception that lets you see the rule. Wikipedia mostly discovered its constitution by living it, and once, at the very start, it set a single line down and grew a community to match.

A rule against rules

The most revealing artifact of the period is also the most paradoxical. Among those first “rules to consider,” one came to be remembered as the earliest of them all, and it instructed people to ignore the rules.

The oldest surviving wording begins, “If rules make you nervous and depressed,” and goes on to tell the intimidated newcomer to set the rules aside and simply get on with editing.

Written as reassurance for people too daunted by procedure to contribute, it was a vaccine against bureaucracy, administered before any bureaucracy existed.

By late 2002 even Sanger had turned against the rule he wrote. He had been its first supporter; now he moved his name to the opponents, deciding the project had outgrown it.

The argument never fully closed, and the page collected wry mottoes for years, among them the suggestion that one should know the rules before breaking them. The rule endured anyway.

The community had decided, early and almost without noticing, that the spirit of a guideline outweighed its letter, and that a contributor improving the encyclopedia held the high ground even when some written page said otherwise.

Policy as scar tissue

For most of 2002 the project’s chief antagonist was its own success. Traffic outgrew the hardware, and through the early summer the article count visibly stalled while volunteers wrestled with server failures the software could not yet handle at scale.

The fix came through a rewrite. A university student named Magnus Manske built a database-backed replacement, deployed in January 2002, that introduced the apparatus editors now take for granted: separate spaces for talk pages and user pages, watchlists, automatic signatures, and a new kind of user called the administrator, empowered to delete pages and block accounts.

For its entire first year, Wikipedia had run with nothing resembling a modern administrator. The role arrived as a concession to necessity. A further rewrite by Lee Daniel Crocker followed, and the engine eventually took the name MediaWiki in 2003.

The original target had been modest by later standards. The founders set their sights on a hundred thousand articles, a mark the English edition reached on 23 January 2003, a mere two years in.

The content rules that now feel eternal were also answers to specific pain. Neutrality had left one question open: what do you do with a view that is published and sincerely held, yet believed by almost no one? The answer hardened across 2003 into two policies, verifiability and no original research, both of them grown out of the neutrality problem.

The no-original-research page, taking shape in December 2003, was little more than a quoted message from the mailing list, laying out a sliding scale: a majority view should be easy to source from standard references; a significant-minority view should come with prominent names attached; and a view held by a tiny minority did not belong in the encyclopedia, whether or not it happened to be true.

That scale traces to a remark by Jimmy Wales, and it did double duty, since the same passage became the cornerstone of the neutrality policy’s rule on undue weight, which adds proportion back into balance.

What now reads as impersonal law began as one paragraph quoted to settle one fight, the encyclopedia’s refusal to become the first place a new theory of physics appeared in print.

The right to leave

Governance matured the same way, under strain. In the first years, serious disputes landed on Wales personally, in the manner of a benevolent monarch who settles quarrels because no one else has the standing to.

By late 2003 there were too many quarrels for one desk. On 4 December 2003 he created two bodies to take the load off it: a Mediation Committee to help editors reach agreement, and an Arbitration Committee to impose binding decisions when agreement failed. The first arbitration panel seated twelve members, divided into three groups of four, with staggered terms meant to run for years.

Wales kept the theoretical power to overrule or dissolve the committee but said he expected to use it almost never, and from that point authority kept flowing downward, from one founder toward a community learning to govern itself. The polity was inching from monarchy toward something with courts.

The deepest check on power was structural, and it had been present from the start. Wikipedia’s text was released under a free license, the GNU Free Documentation License, which meant anyone could take the entire body of work and walk away with it.

The right to fork was a guarantee, built into the license from the beginning, and in February 2002 the Spanish-language community exercised it.

Alarmed by a passing suggestion from Sanger that the site’s commercial parent, Bomis, might one day run advertising, a group led by Edgar Enyedy lifted the Spanish articles wholesale and founded an independent encyclopedia hosted by a university in Seville. The departure left Spanish Wikipedia nearly dormant for a year.

The lesson registered quickly at the center. Advertising became politically untouchable. The Wikimedia Foundation incorporated in Florida in June 2003, and the freedom to leave disciplined the people who stayed.

Naming the bones

By the time the famous summary, The Five Pillars, appeared, the body it described was fully grown. It was written on 4 May 2005 by an editor using the handle Neutrality, in an expansion of an even terser three-point summary that had circulated under the name Trifecta.

The pillars gathered principles already four years old and arranged them as an orientation for newcomers, a map handed to arrivals in a city that had built itself without one.

We tend to picture institutions starting with a charter and then accumulating behavior, the way a building starts with a blueprint.

Wikipedia by and large ran the sequence backward. The conduct came first, shaped by people mostly trying to stay out of one another’s way long enough to write something useful, and the rules were the names attached afterward to patterns that had already set.

The most consequential rules of the open web were written by a handful of people who were not trying to legislate. They were trying to keep the doors open and the damage small, and in the process they discovered a constitution they had been living inside the whole time.