evidence architecture

Three questions that get confused

Most of the disagreement about audio of this kind comes from mixing up three questions that have three different kinds of answer. Fluctara keeps them apart and labels which one a statement is answering.

The three layers

Technical: Can this exact object be identified? Answered by checksums, manifests, render receipts, measured properties, and delivery records. These are verifiable and they are the strongest thing on the site — but they say nothing about a listener.
Research: What does a study support? Answered by a published record that exposes its method, its setup or population, its result, its uncertainty, its limitations, when it expires, and whether it has been corrected.
Personal: What will this do for me? This layer has no general answer. A technical receipt cannot produce one and a study on a population cannot produce one for an individual.

Why the separation is enforced rather than intended

Good intentions do not survive a redesign. The separation is checked by the same automation that publishes the site: a set of claims is refused publication if it lacks the fields that make it assessable, and the deployment gate rejects specific unsafe phrasing outright, including copy that was previously published and later withdrawn.

That is a deliberately unusual arrangement. It means a future editor cannot quietly reintroduce a claim by writing convincing prose, because the gate does not read prose for persuasiveness.

Correction is part of the design

A published record can expire and can be superseded. When something is corrected, the correction is a record rather than a silent edit. If you believe something published here is wrong, write to protoscience@anulum.li and say which statement and why.