Wiki / 3 of 7

Evidence Has an Address

Where a component runs and where it delivers evidence of its work are two different facts. Confusing them means reading an old copy as fresh truth.

A measured case

Six components on one planet read as "broken" on the very machine where they were running flawlessly. Their evidence-a health receipt-was shipped to the control node every fifteen minutes. But the observer opened the same path locally and found a four-day-old copy left behind by an earlier mechanism.

An old copy at the correct address is worse than no copy: absence would have been honest, while this was a confidently false statement about a healthy component.

The rule

An evidence declaration now names the node where the evidence lands. If you are not that node, the path is not read at all: whatever lies there is not evidence.

An error in the node name is caught when the declaration is made. Otherwise, a single typo could turn a mandatory receipt into a permanent excuse that never fails anywhere.

Declared and delivered

The broader principle is that a declaration is a promise; only what has been delivered is evidence. A component that declared a data source and delivered nothing through it is not "working"-it is "promising." Those two words are never rendered the same way in the system.

Site publications now live the same way: each carries a receipt of checks with its own SHA-256 address, and acceptance reads the live artifact and matches it against the declared commit. Four publications in a single day went down this route: source ref, version, live marker - three facts that must agree.

Claim state (self-assessment)holds
Snapshot2026-08-26
What would falsify thisEvidence read from somewhere other than its declared address and accepted, or a declaration made without checking that the recipient exists.
Next stepEvery new evidence declaration undergoes address validation at registration.