Evidence & Verdicts

105 dated documents recording that the method described in the wiki was actually carried out — and what happened when it was.

Why this is a separate section, and why it exists at all. Anyone can publish an architecture page describing a rigorous method. What is expensive to fake is a dated ledger of that method running, kept whether or not the result was flattering. These are those records. They are here rather than in the main index because 105 evidence stubs would bury the two dozen documents that actually explain the system.

What a receipt is

A pre-registration is written before a run, and pins what would count as a pass and what would count as a falsification. Writing it afterwards is the failure mode it exists to prevent — you cannot decide what the evidence meant once you have seen it.

A verdict is the output of an adversarial review in which five separate personas attack a proposed change independently, and a sixth defends the human who will have to operate it. One of the verdicts published here came back FAIL and held a ship gate. That one is worth more than the rest combined, because it is the evidence that the review is real rather than decorative.

A receipt is the record of a specific run: what was measured, on what commit, with what result — including the adverse ones.

Pre-registrations (32)

Written before the run, pinning the pass and falsification criteria in advance.

Adversarial review verdicts (9)

Five personas attacking a change independently, merged into one verdict. Includes at least one FAIL.

Receipts and handoffs (64)

Dated records of individual runs, incidents and session boundaries.