The link you cited is gone
Somebody asks about a position you took two years ago. You find the piece. The source link returns nothing.
That is the good case. The bad case is that it resolves, to a page that has been edited since, with the sentence you quoted no longer in it and no note that anything changed. Your response now reads as though you argued with something nobody said, and the person asking cannot tell whether you misread it or the publisher moved it.
The promise has been getting weaker, and the dates are public. Google retired its cache in 2024, so the reflex “look at the cached copy” stopped working. The Wayback Machine removes pages at an owner’s request and makes no promise about the outcome. By 2026 hundreds of news sites, the New York Times among them, turn away at least one of the Internet Archive’s crawlers. And a study of the web as it stood between 2013 and 2023 found that a quarter of those pages are gone. None of this is malicious and all of it is normal. But an argument anchored to somebody else’s URL is anchored to somebody else’s future decisions. A copy you do not hold is a copy you may not have.
What we keep. Every article we answer is captured: the rendered page, a print-stable PDF, a full-page screenshot, a thumbnail, the readable text, and a provenance manifest with a hash of every artefact, plus a best-effort save to the Internet Archive as a third copy that is not ours. That is an implementation, not an advantage; preserving a page with a hash and a timestamp is an old craft, and a law school gives the software away. What is unusual in this market is the join: the page you answered, the answer you wrote, the person who signed it and the date, in one record, for a communications team. For our own archive the copies sit in write-once storage that refuses deletion and overwriting from the S3 side; the period is our own setting, and a customer’s retention is a contract term set with counsel. We promise no period on this page. running on our own newsroom The record
Where it stands today, said plainly. All thirty of our answered articles have an archive page in both languages, sixty pages. Every capture is dated 6 September 2026, so the archive proves what each page said on that day, not on the day we first answered it. No source has died yet. Nothing watches for it yet: the check that would notice a change or a removal is designed and not built, and today liveness is set when a person checks. The archive has not rescued anything. It will, and we will say so when it does. These are our own numbers. Where your file lives
Where that argument is weak, in the same breath. No source of ours has died yet, so for our own newsroom this is still insurance rather than a rescue. If your beat is stable and your sources are institutional, it may stay that way for years. And some buyers will decide they would rather not hold copies of other people’s pages at all, which is a legitimate answer and one we would rather hear in the first meeting.
What we deliberately do not do. While a source is alive, none of that copy is public. Publicly we are provenance, a thumbnail, a short extract and a live link that passes authority to the publisher, on purpose. The faithful copy of their page is never public, in any state. The full text can appear only after a source is confirmed removed, and only after a person approves it; no scheduled job publishes somebody else’s text on its own. A publisher who asks is moved to provenance-only, without an argument. And we do not retract our own article because its source died: our commentary is our own work, and deleting it would destroy the record the archive exists to build. The law on how long a copy may be kept is not settled, and we would rather say that on this page than in a letter. The record Privacy