The web is losing its own past.

Decay with a half-life

The useful thing about link rot is that it behaves like the other curves on this site. Links do not all fail at some expiry date. They fail at a roughly constant rate, which means a population of links has a half-life: a period after which about half of them no longer resolve. Studies keep landing in a similar region, and the exact figure matters less than the shape.

Drag the years below and watch a set of links age.

What it takes with it

This would be a curiosity if links were decoration. They are not. They are how the web states its evidence. A court opinion citing a source, a paper citing its data, an article citing the document it is reporting on: when those references stop resolving, the claim survives and the proof does not. The text still says the thing. You just cannot check it any more.

Reference rot is the more insidious cousin. The link still works, but the page behind it has changed, so a citation now points at something that no longer says what it was cited for. Nothing looks broken. Everything is quietly wrong. That is exactly the failure from act one: not the record vanishing, but the record being altered without notice.

Where the four meet

A model has no memory, only a transcript, and it can be forged. A person forgets on a curve unless something intervenes on a schedule. A DRAM cell holds nothing at all beyond a few milliseconds without being told again. And the web keeps its links long after it has lost what they pointed at.

Four systems, one problem. The ones that survive are the ones that treat remembering as a thing you do repeatedly, not a thing you did once. Every archive, every refresh cycle, every review is the same move: pay again, now, so the record is still there later.

Every memory system has to forget. The ones that don’t forget on purpose forget by accident.