Why half of every 'useful websites' list is already dead

Social bookmarking was the biggest single casualty. A dozen services competed to be the place where you saved and tagged links, and their entire value came from the number of other people using them. Below a critical mass they became useless, and once they became useless the decline was self-reinforcing. Almost none survive.

The same logic took out the standalone video communities, the social music discovery services and the Q&A sites that were not the biggest Q&A site. In every case the product worked fine; it just could not survive being second.

A large group of the classic entries were Flash applications. When Flash was discontinued they stopped working overnight regardless of whether anyone was still maintaining them. Sites built on a single third-party API — a maps provider, a social network's feed, a music catalogue — went the same way whenever that provider changed terms.

The lesson is not that platforms are bad. It is that a tool inherits the lifespan of the thing it depends on most.

Tools that do one narrow thing, need no network effect and cost almost nothing to run have a remarkable survival rate. A converter, a colour picker, a timer, a regex tester — these have no reason to shut down, because there is nothing to sustain except a static page and a small amount of compute.

The best of them have gotten more robust rather than less: many that used to upload a file now do the work in the browser, which makes them cheaper to run and better for the user at the same time.

Public libraries, university course archives and the big preservation projects have outlasted nearly every commercial entry on the original lists. They are funded differently, they have a mandate rather than a market, and their content gets more valuable with age rather than less.

If you want a bookmark that will still work in a decade, this is the category to pick from.

Prefer tools with no network effect for anything you rely on. Prefer tools that work offline for anything you rely on urgently. Assume anything whose value is a community will either become the dominant one or disappear, with very little middle ground.

And record what a bookmark was for, not just its URL. When it eventually dies, the note is what lets you find the replacement in thirty seconds instead of reconstructing the question from a dead domain name.