Wales, Jimmy
▶ Watch on YouTubeCore ideas
Wales traces Wikipedia's origin to watching the free-software movement and wondering whether that same collaborative, commons-based model could extend beyond code to an encyclopedia. The first attempt, Nupedia, imposed a seven-stage academic review process that produced almost nothing in two years (and even its first published article had to be pulled after the community spotted plagiarism); Wikipedia launched in January 2001 as a low-stakes side experiment using wiki software, and immediately outproduced two years of Nupedia's effort in two weeks, because anyone could just start typing. He recounts early technical accidents that turned out fine — using CamelCase links replaced by square brackets (which German keyboards lack, forcing workarounds that didn't slow German Wikipedia's growth), and choosing human-readable, "interpret liberally" text formats over efficient binary ones so that anyone could learn by viewing a page's source, a philosophy he sees as central to the web's original openness.
On editorial philosophy, Wales frames Wikipedia's core discipline as distinguishing encyclopedic "summary" from raw source material (spinning off full texts into Wikisource) and constantly negotiating "notability" — which he considers a poorly chosen but functionally necessary concept, preferring to think of it as verifiability: is there enough reliable source material to write something responsibly? He is especially careful about biographies of living persons, citing categories like "BLP1E" (a person notable for only one event, such as a crime victim) as a "danger zone" where Wikipedia should generally not have a standalone article, and describing "undue weight" (e.g., a minor DUI arrest dominating a decades-long local politician's biography) as a recurring problem the community actively polices. He argues source quality matters enormously — citing Wikipedia's decision to "deprecate" (not ban) the Daily Mail's online arm as an example of steering editors toward better sourcing rather than centrally banning viewpoints — and stresses that "community health" (how editors react to good-faith correction versus defensiveness) matters more than any single content dispute.
Asked directly whether Wikipedia has a left-leaning bias, Wales says he doesn't see it as a systemic problem, arguing genuine bias more often shows up in obscure, non-political topics (he cites unusually glowing anime articles, written almost entirely by fans) than in high-traffic political ones, where he believes the community works hard to represent competing "meaningfully important" perspectives (contrasted with false-balance cases like flat-earth or "moon made of cheese," which get almost no serious counter-treatment). He walks through the COVID mask-efficacy article as a case he considers well-handled — noting evolving, mixed evidence — and separately criticizes public-health authorities' early "don't buy masks" messaging followed by a hard reversal as corrosive to public trust, a theme he extends to a broader argument (echoing Lex's own view) that authorities scaring the public into "correct" behavior with overstated claims damages long-run trust in science and institutions more than honest uncertainty would. He rebuts specific viral accusations of Wikipedia bias or censorship in detail — including claims that Wikipedia altered the definition of "recession" to help the Biden administration (he says only a paragraph was reordered, prompting a temporary wave of trolling) and that Wikipedia suppressed the "Twitter Files" article (he says a routine, near-instantly rejected single-editor deletion nomination was not suppression) — framing both as cases where an ordinary editorial process was misreported as ideological censorship.
On large language models, Wales sees ChatGPT-style tools as broadly complementary to Wikipedia's mission (its content is freely licensed specifically so it can be reused), while flagging hallucination as the central unsolved problem — describing a personal experiment where GPT-4 fabricated plausible-sounding but nonexistent quotes when asked about athlete comparisons — and proposing that Wikipedia's own norm of visibly flagging "the neutrality of this section is disputed" is a model of honest uncertainty-signaling that AI systems (and, he jokes, mainstream newspapers) could adopt. He describes internal experiments feeding an article and its sources to an LLM to auto-flag inconsistencies, and using AI to summarize talk-page controversy as a promising way to give readers a sense of how contested a given passage really is. On business model, Wales frames Wikipedia's ad-free, small-donor-funded structure as a deliberate (originally more aesthetic than strategic) choice to prevent revenue incentives from corrupting neutrality and content prioritization, contrasting it with how advertising-driven platforms like Facebook and Twitter structurally reward outrage and engagement over long-term wellbeing — a dynamic he illustrates with a "sweet grandma vs. racist uncle" analogy for how algorithmic amplification elevates whichever family member gets the most heated replies. On government pressure, Wales states Wikipedia has never complied with censorship demands anywhere in the world and never will, while distinguishing that hard line from legitimate dialogue (e.g., discussing how Wikipedia's editorial process works with health agencies or foreign governments) that never involves altering content to appease anyone. He closes with reflections on the community's demographic skew (Wikipedia's editor base is heavily male and tech-oriented, which he says produces real content gaps — better-developed articles on male novelists' award-winning books than equally-acclaimed female novelists', for instance — driven by what people choose to write about rather than deliberate bias) and his belief that Wikipedia's endowment and financially conservative structure position the foundation to persist for a century or more.
Excerpts
On why Wikipedia treats biographies of people notable for only one event as a "danger zone":
“One of the areas this comes up quite often is what we call BLP1E... The type of example would be a victim of a crime, so someone who's a victim of a famous serial killer, but about whom really not much is known. They weren't a public person, they're just a victim of a crime, we really shouldn't have an article about that person... how can you write a biography about someone you don't know much about?”
On why he believes Wikipedia's political bias is overstated, illustrated by an accusation over the definition of "recession":
“The accusation was often quite ridiculous and extreme, which is, under pressure from the Biden administration Wikipedia changed the definition of recession to make Biden look good... What happened was, one editor thought the article needed restructuring... the only thing that really changed was moving that from the lead... to further down. And then news stories appeared saying, 'Wikipedia has changed the definition of recession.'”
On Wikipedia's refusal to comply with government censorship demands:
“We've never bowed down to government pressure anywhere in the world, and we never will... If they threaten to block, well, knock yourself out, you're going to lose Wikipedia. And that's been very successful for us as a strategy because governments know they can't just casually threaten to block Wikipedia... and we're going to cave in immediately.”