This is a composite interview. The quotes in it are drawn from public talks, interviews and writing by people who do this work; no single archivist sat for it, though plenty have lived it.
The post always reads the same way. After careful consideration. An incredible journey. We’re thrilled to announce our next chapter. Somewhere around the third paragraph, past the gratitude and the photo of a sunset, there is a date — the day the servers go dark. When Posterous published that post on 15 February 2013, it gave its fifteen million users ten weeks. The podcast hosts Podango and MyPodcast, when their turns came, gave four or five days. And the moment such a post goes live, a very particular kind of person springs into action: scripts are written, hard drives are cleared, bandwidth is donated by strangers. The people who do this call it rescue.
Most of us would call it archiving, which undersells it — archiving suggests a quiet basement, and this is closer to carrying furniture out of a building the demolition crew has already started on. When Yahoo announced in 2009 that it was closing GeoCities, millions of hand-built homepages — the first place a generation of people ever existed online — were scheduled for erasure. Archive Team, a loose volunteer collective led by Jason Scott, spent from April to October copying. They saved roughly a terabyte of millions of pages: a heroic sliver of a service that held some nine terabytes at its peak. The rest went down with the ship.
Here is the part that makes this more than a story about digital hoarding. The rescue is, legally, the same act as the theft. Copying a website wholesale is precisely the thing copyright law exists to restrict; deleting the same website is a business decision. Which means the cultural record of the last thirty years is being kept largely by people working in a grey zone — unpaid, unofficial, tolerated rather than authorized, and always against a countdown. Scott has described the operating principle with admirable economy: “Archive Team doesn’t ask. It takes.”
The default is disappearance
The strange thing about link rot is that we keep treating it as decay — as though the web were an old building slowly crumbling. It isn’t. Disappearance is the web’s resting state. When Pew Research Center sampled the recent web in 2024, it found that a quarter of everything that existed online between 2013 and 2023 was already inaccessible: 38 percent of pages from 2013 were gone, and even 8 percent of pages from 2023 had vanished within months of their birth. The supposedly serious internet fares no better. Twenty-three percent of news webpages and 21 percent of government webpages contain at least one broken link, and more than half of Wikipedia articles cite at least one source that no longer exists. Nearly one in five tweets disappears from public view within three months. The internet has no persistence layer. Persistence is an aftermarket part, fitted by hobbyists.
Ask the hobbyists why they bother and the answers are disarmingly sentimental. Brewster Kahle, who founded the Internet Archive, has talked about amateurs’ “emotional investment” — the recognition that somebody’s glitter-texted GeoCities page about their cat was, to that somebody, a life’s work. Scott talks about the notices themselves: Poetry.com held around fourteen million poems and gave about a month’s warning — then spent its final weeks actively blocking Archive Team from downloading the archive. Picture that: a dying company devoting some of its last engineering effort to making sure the poems died with it. Scott’s read on the legal climate is blunt. “We all know that this country is a little psychotic about copyright,” he told On the Media. “We’re not selling what we’re putting here.”
That is the grey zone’s whole defense: the archive is a gift, not a product. It has not always been enough. In 2020, four major publishers sued the Internet Archive over its book-lending program; a federal court ruled against the Archive in 2023, an appeals court upheld the ruling in September 2024, and by December the Archive had given up appealing. A separate lawsuit from music labels over its digitization of old records could cost it $400 million. Whatever you think of the Archive’s legal theories, the docket tells you something about priorities: the fastest way to get sued on the internet is to keep something.
Now consider what happens when the opposite occurs. In 2019, MySpace announced that a server migration had made all photos, videos and audio uploaded more than three years earlier unavailable. The technologist Andy Baio estimated the loss at around fifty million tracks by fourteen million artists — and publicly questioned how accidental such a loss could really be. There was no lawsuit. There was no injunction. There was a brief news cycle and then nothing. Fourteen million artists’ work evaporated as a line item, while elsewhere the courts were carefully dismantling a library for lending books too enthusiastically.
The wrong side of the ledger
The institutions that are supposed to do this officially are arriving, slowly. The UK extended legal deposit to online works in 2013, meaning its deposit libraries may now lawfully copy websites for preservation — official memory moving at the speed of legislation, while platforms die at the speed of a blog post. The informal net still catches more. Mark Graham of the Internet Archive estimates that of defunct links, roughly two-thirds are recoverable through the Wayback Machine. Two-thirds sounds reassuring until you turn it around: a third of the dead web is simply gone, recoverable by no one.
And this is the asymmetry running under everything. Deleting is free, instant and legally protected — celebrated, even, as “sunsetting,” as though fifteen years of human expression were a lawn that needed mowing. Keeping is expensive, perpetual and one court filing away from ruin. Every incentive we have built sits on the wrong side of that ledger. The delete button is the only feature of the internet that never breaks, never rots, never returns a 404. Somewhere this week, a product team is drafting the next incredible-journey post, and somewhere a volunteer is checking the free space on a hard drive. Only one of them is racing. The other one owns the clock.