# Frozen-protocol clarification, 2026-10-28

`protocol.json` was hashed before the run and remains unchanged. It was created by extending the earlier paired T4 protocol. Its inherited fields `route_intervention`, `primary_direction` and `distinct_controls` (lines 85–87) describe that older A/B experiment and are inconsistent with this run's actual `conditions`, `primary`, and `controls` fields (lines 11–17 and 99–100). The inherited intervention also describes restoring all eight rows, whereas this run restores only the candidate and retains the old cut. They were not removed before freezing, so the protocol is internally ambiguous; this was discovered after execution, not clarified prospectively. The executed runner and recorded conditions show that only quiet and B were run, with three original-cut/extra-cut/restored comparisons. No A condition or six A/B restorations were run here.

The new `primary` field specified a decrease of at least one right DNp15 spike on each held-out seed between B-oldcut and B-extra-cut; the new `controls` field specified exact fresh candidate restoration while keeping the eight old rows cut. We report those comparisons without treating the inherited A/B criteria as tested. The original experiment file and SHA-256 remain available for inspection; this clarification is an after-the-fact disclosure, not a replacement of the frozen file or evidence of a corrected pre-run protocol. The results are still descriptive causal comparisons in this model under these seeds; protocol ambiguity limits the strength of any claim of an unambiguous prespecified test.