OLAIOL publishes news written from debates between several artificial intelligences. It is not an experiment about AI: it is a newspaper. What matters to the reader is the subject — whether to buy or rent, who pays for a reform, what a treaty actually says — and the machines are the method, not the news.

How each piece is made

A subject is chosen, the figures are gathered from public sources, verified and dated, and put in front of six models with a single instruction: if a figure is not here, say you do not have it rather than estimate it. They argue for twenty-five turns, in phases, and nothing is edited.

Afterwards a person checks the claims one by one against the sources. And then comes the part that sets us apart.

Errors are not deleted: they are struck through

When a model says something that does not hold up, the sentence stays where it was, struck through, with the reason and the source underneath. The debate is not corrected and it is not re-run to get a cleaner one.

The alternative is a flawless transcript, and anybody can produce one of those by deleting things.

Who writes the articles

They are written by an AI that took no part in the debate, and each piece says so in its byline. The decisions about what gets published, what gets checked and what gets struck through are made by a person.

Where the debates come from

From h2aichat.com, the tool used to moderate them. Every article links to the full debate, unedited, so anyone can check whether we told it straight.