The post comes from the author of 0.mk, a URL shortener that originally ran from 2009 to 2014 and has now been revived. Using a recovered database backup, they crawled 657,607 historical shortened links and found that 76.7 percent of the crawlable targets no longer returned a loading page. That is a generous number. Pages blocked by 403 and 429 responses may still exist, while pages that return 2xx or 3xx still count as loading even if they now land on parked domains, login walls, or removed-content notices. A big outlier from one spammy account distorted the 2011 data, but removing it did not change the broader pattern. The post also doubles as a test of whether AI can now do enough of the engineering, abuse review, spam filtering, and support work to make a tiny service like this viable again.
People mostly took the decay number as believable, but the sharper point was that 23.3 percent “alive” is only an upper bound. An
HTTP response does not mean the original thing is still there in any useful sense. The thread also pushed back on the article’s framing of 2009 to 2014 as the “old web.” For many readers, that period was already
Web 2.0 or the “old social web.” What they mean by the old web is earlier and more decentralized. Think personal homepages, niche forums,
IRC, blogs, web directories, and community-run spaces before Facebook, Reddit, Discord, and algorithmic feeds pulled activity into a few giant platforms.
Where the conversation landed was less nostalgia for GIF-heavy Geocities pages than a loss of ownership, discoverability, and permanence. Older communities often had their own domains and one or two technically inclined people running the stack. That meant local norms, ugly design, and occasional chaos, but also real autonomy. Today many communities still exist, yet they live on corporate platforms that control ranking, moderation boundaries, login walls, scraping access, and monetization. Several commenters argued that AI is making this worse on both sides at once. It floods search and content discovery with slop, and it triggers anti-bot defenses that make the web slower, less open, and harder to archive. That leaves a web that may still technically contain the same niches, but in a shallower, more fragile, and more controlled form.
A smaller but recurring thread said the old web is not really recoverable as a technology stack because what people miss is the social moment. Early web culture felt magical partly because the medium itself was new and self-selecting. Communities die when people leave, not only when platforms change. Even so, the practical consensus was clear. If you care about durable knowledge or durable communities, static pages, owned domains, open protocols, and archives still beat feeds, apps, and rented audiences.
0.mk