The History of YouTube

In February 2005, three former colleagues from an online payments company registered a domain for a website that was supposed to help people find dates. The pitch was that members would upload short videos describing themselves, and other members would watch and get in touch. Chad Hurley, Steve Chen and Jawed Karim built it above a pizzeria and a Japanese restaurant in San Mateo, California, and then discovered the flaw that would make them rich: almost nobody uploaded a dating video, and the few who did were ignored. What people wanted to upload was everything else.

So they took the dating part out. On 23 April 2005, Karim stood in front of the elephant enclosure at the San Diego Zoo while a friend filmed him on a handheld camera, and delivered eighteen seconds of nothing much at all — a remark about how the elephants had really, really long trunks. Me at the Zoo was the first video uploaded to the site. It is still there. It is a genuinely strange artefact to revisit, because it contains no trace of the thing it started: no title card, no intro, no channel branding, no call to subscribe, no format. It is just a young man in a zoo, being filmed. Every convention of the medium that followed had to be invented by somebody, from scratch, in public.

The Problem Nobody Else Had Solved

Video on the web in 2005 was miserable. Clips arrived as downloads in half a dozen mutually hostile formats, each demanding its own player, each capable of freezing a computer. Bandwidth was expensive and asymmetric — home connections were built to receive, not to send. Anyone who wanted to put a video online had to be, in effect, a small systems administrator.

What the new site did was remove every one of those decisions from the person uploading. It accepted whatever file you had, transcoded it server-side into a single format that played inside the browser through the plug-in nearly everyone already had, and gave you back a URL and a snippet of code you could paste anywhere. That last part mattered more than the rest combined. The embeddable player meant a video did not have to live on the site to be watched from the site: it could sit inside somebody else's blog post, forum thread or social profile, and every one of those pages became a doorway. The service grew because other people's websites carried it.

Money followed the traffic. Sequoia Capital put in $3.5 million in November 2005 and more the following spring; the site left beta and launched properly in December 2005. And then, in October 2006 — twenty months after the domain was registered, with a company of roughly sixty-five people — Google agreed to buy it for $1.65 billion in stock. At the time the number was widely treated as evidence of a second dot-com mania. It is one of the more comprehensively settled arguments in the history of technology.

The Copyright Wars

The thing that made it grow was also the thing that nearly killed it. If anyone can upload anything, then anyone will, and a great deal of what went up in those early years belonged to somebody else.

The tipping point was a sketch. Lazy Sunday, a two-and-a-half-minute rap about cupcakes and a matinee of The Chronicles of Narnia, aired on late-night American television in December 2005, was uploaded within hours, and was watched millions of times by people who had not seen the broadcast and had no way to see it again. It was the clearest possible demonstration that the audience for a piece of television was not the same thing as the audience for a television slot. In February 2006 the network that owned it asked for the clip to be taken down, and it was.

The lawsuit came a year later. In March 2007 the media conglomerate Viacom sued for $1 billion, arguing that the site was built on infringement and knew it. The case ran for seven years, produced a district court ruling in the platform's favour in 2010 on the strength of American safe-harbour law, survived appeal, and was finally settled in March 2014 on terms neither side disclosed.

The engineering answer arrived before the legal one. From 2007 the platform began fingerprinting uploaded audio and video against reference files supplied by rights holders, and — critically — gave those rights holders a third option beyond allow and remove: monetise. A studio could let a fan's upload stand and take the advertising revenue from it. That single design decision converted a war into a business relationship, and it quietly defined the terms on which nearly all borrowed footage has appeared online ever since. It also handed enormous, largely unaccountable power to whoever holds the reference file, which is an argument that has never stopped.

The Invention of the Job

In May 2007 the platform announced a partner programme: a share of the advertising revenue, paid to the people making the videos. It is difficult to overstate how strange this was. Nobody had ever been paid a proportion of the advertising against a thing they filmed in their bedroom and uploaded themselves. Television paid people through commissions, salaries and residuals, all of which required an institution to say yes first. This required nothing but an audience.

Within a few years there was an occupation that had not previously existed, with no name that anyone was happy with. The people doing it had no agents at first, no union, no standard contract and no idea what the rates were. They had a direct line to an audience and a monthly payment whose size they could not predict and could not appeal.

The economics shaped the work with an almost brutal directness. When the upload limit was ten minutes, videos were under ten minutes. When ranking rewarded raw view counts, thumbnails and titles inflated into the shouting style everyone now recognises. In 2012 the platform changed the signal it optimised for from views to watch time — how long people actually stayed — and the medium visibly changed shape within about eighteen months. Videos got longer, then much longer. The mid-length essay became viable. The two-hour conversation became viable. A format that had been defined by brevity because the machinery demanded brevity turned out not to have been about brevity at all.

What The Medium Actually Invented

The forms came from the audience side, not from any commissioning process, and most of them have no real television ancestor.

The vlog — a person addressing a camera about their own life, daily, indefinitely — is not a documentary and not a talk show. It has no episode structure, no arc and frequently no subject. It works because continuity of presence is itself the product.

The let's play and its descendants took an activity that was previously private and made watching someone else do it the entire point, which is the same discovery that turned live gaming into a global business.

The video essay grew out of the watch-time era into something genuinely new: long, referential, densely illustrated criticism, made by people with no institutional platform, running to lengths no broadcaster would ever have commissioned.

And the reaction, the challenge, the haul, the unboxing, the explainer — each one an entire genre with its own conventions, its own stars and its own economics, none of which existed before the medium that produced them.

Meanwhile the reverse experiment failed instructively. In 2011 the platform put roughly $100 million into funding professionally produced "original channels", on the reasonable theory that better production values would attract better audiences. Most of them did not work. What the money bought was television made slightly cheaper and put in the wrong place. The channels that thrived alongside them were made by people who understood that the medium's native form was not a smaller television programme.

The same lesson arrived from the business side. Between roughly 2011 and 2016 a wave of multi-channel networks signed up thousands of creators, promising sales, production support and scale, and took a cut. Most were sold, merged or collapsed. The intermediary layer that television requires turned out to be, in this medium, largely optional — and the creators who survived best were the ones who had built the direct relationship rather than outsourcing it.

The Fragility Underneath

The direct relationship had a hole in it: the money came from advertisers, and advertisers could leave.

In early 2017, brands discovered their advertisements running against extremist and abusive content and pulled their spending. The platform's response was broad and blunt — sweeping automated demonetisation that caught enormous quantities of ordinary work alongside the material it was aimed at. Creators who had built a livelihood over years watched it fall by half in a week for reasons no human had explained to them and no human would review. The episode acquired a nickname, and it recurred.

What came out of it is the structure the medium runs on now: almost nobody serious depends on advertising alone. Direct membership, crowdfunding, the read-aloud sponsor segment, merchandise, live events, the paid subscription tier the platform launched in 2015 — the modern creator business is a portfolio, assembled deliberately, because the lesson of 2017 was that a single revenue source controlled by somebody else is not a business, it is a tenancy.

Then in 2020 a rival made vertical short video the dominant format among younger audiences, and the platform answered with Shorts, rolled out through 2020 and 2021. That fight is not over, and it is producing its own medium — one with different economics, different lengths and different conventions again. Which is rather the point: this has never been one stable thing.

Why It Is Not Television

Which brings us to the reason this history sits on a film and television site at all.

A hit series made for this platform can run for a decade, employ a real crew, book guests any network would want, win awards and be recognised on sight by more people than most prime-time shows. Hot Ones, launched in 2015 by the food publisher First We Feast and hosted by Sean Evans, is a celebrity interview show whose entire structural gimmick — the guest answers increasingly serious questions while eating increasingly punishing chicken wings — has produced some of the most-watched celebrity interviews of the last decade. Nobody commissioned it. No network ordered a pilot. No slot existed for it to fill.

That is the difference, and it is not a difference of budget or quality. It is structural, and it runs through every part of how the work is made:

Television is commissioned — an institution decides a thing should exist and pays for it before an audience is proven. This is published — the thing exists first, and the audience decides afterwards whether it continues.

From that one inversion, everything else follows. There is no season order, so there is no season: shows run continuously, pause when the maker pauses, and resume. There is no slot, so runtime is whatever the piece needs — a fourteen-minute episode and a three-hour one can belong to the same series without anyone treating that as a format change. There is no network standards department, so the editorial line is the maker's own, with the audience and the advertisers as the only real check. There is no cancellation in the television sense, because there is no one to cancel it; series simply stop, often without an announcement, which is why "is it still going?" is a genuinely hard question here in a way it is not for a broadcast show.

The credits are different too. A television drama's crew list runs to hundreds. A long-running series in this medium may be four people, or two, or one — and the one is very often the writer, the director, the presenter, the editor and the business at the same time. That is not a smaller version of a television job. It is a different job.

And the relationship with the audience is direct in a way broadcasting never was. There is no affiliate, no carrier, no schedule and no region: the audience is the customer, the distributor and the comment thread, all at once.

Call all of that television and you have not simplified anything — you have thrown away the only information that explains why the thing looks and behaves the way it does. A twenty-year-old medium with its own economics, its own formats, its own failure modes and its own history deserves its own name.

That is why, here, it has one.