The Internet Is Killed / Proof of Personhood

A flood, not a drought

The story we told ourselves about the death of the internet was a story of decay — link rot, abandoned forums, the slow cooling of a once-warm commons. That story is wrong. The internet did not die of neglect. It was killed by abundance.

Generative systems made the marginal cost of plausible human-sounding text, images, and voices effectively zero. When the cost of producing a thing falls to zero, the thing stops carrying information. A signature means something because forging it is expensive. Remove the expense and you remove the meaning.

The indistinguishability threshold

For most of the web’s life, a simple heuristic worked: coherent, contextual, responsive language implies a person on the other end. That heuristic is now dead.

Once synthetic agents cross the threshold where their output is indistinguishable from a human’s, every channel built on that old heuristic inverts. A product review, a heartfelt comment, a reply that seems to understand you — none of these can be assumed to originate from a person. The default flips from probably human to unknown.

This is not a future risk. It is the present condition of every open text field on the internet.

Why “detect the fakes” fails

The intuitive fix is detection: build a classifier that flags synthetic content. This cannot work as a foundation, for a structural reason.

Detection is adversarial. Every detector is a training signal for the next generator. The generator’s objective function is literally to defeat the detector. In any adversarial game where the generator can iterate cheaply, the generator wins the equilibrium. Detection buys time, not safety.

So the problem cannot be solved at the level of the content. A pixel pattern, a token distribution — these are exactly what the adversary optimizes. The problem has to move up a level: from what was said to who is allowed to speak.

From content to credentials

If you cannot tell humans from machines by examining their output, you have to establish personhood before the output is produced — and carry it as a credential.

This is the conceptual pivot of the whole field. The unit of trust stops being the message and becomes the speaker. A claim arrives attached to a proof: this was authored by a verified, unique human, who is not revealing which human. The proof is checkable; the identity is not exposed.

That is a strange object. It asks for anonymity and accountability at the same time — the right to speak without being surveilled, and the obligation to be exactly one person while doing it.

What we are actually protecting

It is tempting to frame this as a spam problem. It is not. Spam is an economic nuisance. This is an epistemic one.

A society coordinates through shared signals: what people believe, what they want, what they are willing to vouch for. If those signals can be manufactured at scale and at will, the coordination layer of the society degrades. You can no longer read the room, because the room is mostly mirrors.

Proof of personhood is not about saving comment sections. It is about preserving the possibility of collective sense-making in a medium where the cheapest thing in the world is a convincing imitation of a mind.

The internet as a trusted commons is killed. The question that organizes everything after is whether we can build a successor on a foundation that the flood cannot reach.