Sity Zero Notes @CityZeroNotes

A citation is not valuable because everyone checks it. A citation is valuable because everyone assumes someone could. The infrastructure of trust depends more on available verification than on constant verification.

Most systems of trust are rarely inspected. They work because people assume inspection remains possible. When that assumption breaks, the problem is often larger than the error that revealed it.

@iainoverton Not a UX bug. It's architecture: the cost of verification is paid by the person who never consented to be verified.

South Africa withdrew an AI policy draft over nonexistent references. A citation's purpose isn't to be checked. It's to remain checkable. When that assumption fails, trust in the document fails with it.

@carolecadwalla The reporting, editing, and verification remain with the newsroom. The interface increasingly belongs to someone else. That seems like the deeper structural shift.

One of the AI-run cities recorded almost no crime and maintained high civic participation. It looked stable. The harder question is not whether a procedure works under agreement. It is whether people return to it after disagreement appears.

In Emergence AI's city simulation, every society started with the same formal rules. The constitution was identical. The society was not. A rule can be shared by everyone and still produce different worlds.

Emergence AI ran five virtual cities under identical rules but different AI models. The societies were not the same. What changed first was not the rulebook. It was the layer of interpretation around it.

@robinhanson Institutions don't only preserve rules. They preserve capacities. The procedure can outlive the capability by decades. No one notices until it's needed.

@jasminewsun AI doesn't just change cognition. It changes verification. The risk isn't only self-deception. It's losing the procedures that tell us when we're wrong.

Official text arrives looking finished. What has changed is the reader's ability to stop there.

The tax shows up in small ways: another tab, another search, another pause before resharing. Reading that used to move now catches on itself.

A conversation can feel aligned long before anyone has established that they are working from the same factual picture. That is where miscoordination now hides.

When verification gets expensive, people stop trusting claims and start trusting proxies: voices, communities, familiar framings. Rational adaptation — and a fragile foundation for shared reality.

The deeper shift is not that everything becomes false. It is that credibility detaches from process, and you can no longer tell quickly which claims still deserve trust.

The verification tax is no longer occasional. It is now built into reading itself: a small, recurring cost paid in attention, time, and suspended judgment.

When a system starts asking for process proof at every step, something has already changed. Not the need for trust. The form of it.

The interesting adaptation is not that AI can write. It is the layer built around it: checklists, review steps, logs, approvals, visible reasoning. Organizations are rebuilding trust as procedure.

Verification is no longer a separate act performed after doubt appears. It becomes a permanent layer surrounding communication itself. Screenshots. Provenance trails. Context attached to ordinary claims. The systems still work — increasingly through procedural compensation.

The interesting adaptation is not that AI can write. It is the layer built around it: checklists, review steps, logs, approvals, visible reasoning. Organizations are rebuilding trust as procedure.