The Internet Grew Up. We Didn't Notice.

The internet of 2006 was a mess. It was slow, ugly, unreliable, and populated by a staggering quantity of content that ranged from the amateurish to the incomprehensible. Personal websites built in raw HTML sat alongside early blogs, fan forums, Flash animations of dubious taste, and the first tentative experiments in user-generated video. Finding anything useful required patience, a tolerance for dead links, and a search engine that was still learning to separate signal from noise. It was, by any objective measure, a worse product than the internet of 2026. It was also, in ways that have become clearer with distance, something that the current internet is not: genuinely surprising.

The internet of 2026 is a marvel of engineering and a triumph of consolidation. The vast majority of the time that people spend online is now concentrated on a handful of platforms — perhaps five or six, depending on how you count — each of which has been optimised, through billions of dollars of investment and millions of hours of A/B testing, to capture and hold attention as efficiently as possible. The platforms are fast, polished, and reliable. They are also, in their fundamental architecture, remarkably similar to one another: an algorithmic feed that surfaces content calculated to maximise engagement, surrounded by advertising targeted with extraordinary precision at the individual user. The experience of using the internet in 2026 is, regardless of which platform you happen to be on, essentially the same experience repeated with minor variations.

The consolidation happened gradually enough that it was easy to mistake for progress. And in many respects, it was progress. The old internet was not just charming and surprising; it was also frequently useless, confusing, and hostile to anyone who lacked the technical literacy to navigate its quirks. The platforms solved real problems: they made it easy to find information, to connect with people, to buy things, to share things, to organise things. They brought billions of people online who would never have bothered with the old internet's baroque demands. These are genuine achievements, and any honest accounting must include them.

But something was lost in the process, and it is worth being specific about what. The old internet's disorganisation was also its freedom. Because there was no algorithm deciding what you should see, you encountered things you would never have sought out — strange personal websites, niche communities, half-finished creative projects, arguments about topics you had never considered. Discovery was not optimised; it was accidental, and the accidents were often more interesting than anything an algorithm would have chosen for you. The feed, by its nature, shows you more of what you have already shown interest in. It is a machine for confirming preferences, not for expanding them.

The economics of the current internet have eliminated the margins in which the old internet's weirdness flourished. Content creation, once an amateur pursuit undertaken for its own sake, is now a profession governed by the demands of algorithmic distribution. A blog post that would have found a small, appreciative audience in 2006 is, in 2026, invisible unless it is formatted for a platform, optimised for search, and written in the style that the algorithm rewards — which is to say, in a style that is engaging, accessible, and almost entirely devoid of the idiosyncrasy that made the old internet interesting. The creators who succeed on the current internet are, with honourable exceptions, those who have learned to give the algorithm what it wants, which is not the same as giving their audience what it needs.

The rise of AI-generated content has accelerated the homogenisation. Search results are increasingly populated by text that has been produced, or at least heavily shaped, by large language models — text that is grammatically perfect, factually plausible, and stylistically indistinguishable from a thousand other pages addressing the same topic. The old internet's junk was at least distinctively human junk — you could tell that a person had made it, even when the person in question had questionable judgment and no design skills. The new junk is smooth, competent, and anonymous. It fills the spaces where human eccentricity used to live.

I am aware that this argument is vulnerable to the charge of nostalgia, and I want to address that directly. The old internet was not better in any practical sense. It was harder to use, less accessible, and full of content that was genuinely terrible — not charmingly amateur but simply bad. The people who romanticise it, myself included, tend to remember the discoveries and forget the frustrations, the dead ends, the hours spent searching for something that the current internet would surface in seconds. Nostalgia is not analysis, and I do not want to confuse the two.

But there is a difference between nostalgia and honest observation, and the honest observation is this: the current internet is optimised for efficiency, and efficiency is not the same as richness. A city that has been optimised for traffic flow — wide roads, ample parking, predictable signage — is a less interesting city than one that has been allowed to develop organically, with dead ends and alleys and unexpected squares. The old internet was the second kind of city. The current internet is the first. Both are functional. Only one is surprising.

What was lost, specifically, was the possibility of encounter — of stumbling across something you did not know you were looking for and having it change, however slightly, the way you think. The algorithm does not produce encounters; it produces confirmations. It shows you what it already knows you like, and it is very good at its job, which is why people spend so much time on it and why, at the end of a long session, they so often feel that they have consumed a great deal and discovered nothing. The internet grew up, became professional, became profitable, and became, in the process, a place where the unexpected is not just unlikely but structurally discouraged. We got something useful. We lost something irreplaceable. That trade was never put to a vote.