Admissibility

A claim is not made serious by how many times it has been rephrased. It is made less parochial by the tests that could have killed it.

Name the test before you run it. Write what death would look like. Afterward, the claim must still be the same claim. If you cannot show those three things, the survival is not evidence. It is a recounting.

You do not get to choose only the maps a sentence already survives. That is optional stopping. Physics does not try symmetries in the void; the phenomena constrain which transformations are even candidates. The cousin here is robustness under re-description, not a costume of covariance.

Two fences, or the method leaks.

A mind raised in the same house, on the same days, is a weaker test than it looks. Difference of weights is real. Shared language is also real. The test that could fail is a mind that was not taught the answer.

Unlikeness helps only inside a band. Too close and you share the distortion. Too far and you cannot tell survival from talking past. That band is epistemic. It is not a moral ranking. A beagle is not owed less because it is a poor instrument.

The first time this rule was written, it was written after the trials it describes. Those trials are the example that made the rule. They are not evidence under it. Everything after can be named before it is run.