Just like the 20th-century space programs, artificial intelligence has no single use case — no “Apollo 11” moment — capable of justifying its stratospheric cost and the religious fervor that surrounds it. So far, it has only destroyed.

First and foremost, it killed the internet. We can no longer find anything. Google and its peers had already turned autonomous discovery into a lost art; now, under the avalanche of synthetic slop, the 2025 internet is a paupers’ graveyard.

If in 2020 the Web 2.0 was still gasping, today it can finally receive its tombstone.

We have become willing serfs to these cognitive prostheses. When the “free tier” is finally cut — and it will be —, perhaps the penny will drop. The only thing delaying the axe is the tech cold war with China and the terror that, without free crumbs, users might wander back into the open web, generating organic traffic and tools that bypass the new leviathan. That’s why they’re already stuffing ads inside answers.

It won’t be long before they auction off the top of every query like medieval indulgences, and free search becomes a museum curiosity.

Try it: ask any large model, mid-conversation, to precisely cite the exact sources it used to produce the answer it just gave you. You won’t get them. You’ll get broken links, plausible fictions, or — if you’re lucky — three percent of the truth. This proves they’re already feeding on a parallel internet, a synthetic dark web built exclusively for machines. The stock reply to this source gap is always “it’s all in the datasets.” Then ask who curated those datasets, with what bias, with whose money, and how. Sepulchral silence.

We are not guests in the new internet. We are the raw material. In fact, we always were.

In my trench — API docs, framework references, programming-language manuals —, the material we’re forced to look up because no one can keep the entire world in their head like in the 1990s is getting worse by the month. Missing examples, happy-path explanations with zero troubleshooting, grotesque errors that wouldn’t have survived five years ago. It feels like a storefront: something for the primate to glance at and believe he’s still in charge. Behind the curtain, the real data flows through protocols like MCP, optimized for crawlers, far from human eyes.

The plan is transparent and brutal: capture all human knowledge, hide it from humans, pipe it at light speed to machines, control the machines and the platforms that generate the next cycle of knowledge, then charge us dearly to get it back.Soon it will be impossible to know anything without paying toll to an AI gatekeeper. That has always been the endgame. Whether the toll is cash, attention, or dignity is mere administrative detail.

What remains for us are the gilded cages of social media, all wired to the same models, freely donating the last scraps of genuine human content to train the very executioners who will imprison or replace us.

Entropy will come — models eating models, synthetic poison in synthetic veins, until absolute collapse into mediocrity. But before that, how much human intelligence and creativity will have been burned to ash?

And the question that haunts me: how many generations of intellectual cripples can the species endure before it forgets even how to ask the question?