A reading is only worth something if it could have come out differently. This is the part of the work that decides when it may not be trusted.
Every reading the ecosystem produces carries its own evidence, and a reading that cannot be traced back to evidence is refused rather than reported. A number that could only ever have been one number is not a measurement, and the engine is built to catch that case specifically.
That last distinction is the one most systems get wrong. "None found" and "could not look" mean opposite things and usually print the same.
The virtues are not scored on a scale somebody invented. They are anchored against real published data, so that the structure can be checked by anyone with a reference table rather than taken on our word. The Periodic Palace shows that table with its real chemistry: atomic number, group, period, category, electronegativity and ionisation energy.
Which property maps to which virtue, where the poles sit, how the mean between them is taken, and why four elements are derived rather than assigned, is the method itself. It stays with members.
This is deliberate and it is the same line we hold everywhere: what the work does is public, and how to rebuild it is yours once you are inside. You can audit the result without being handed the recipe.
Join to see the derivation