An experiment that decided nothing: a preregistered run below its own resolution threshold

Work that fails to resolve a question is rarely published, and when it is, the failure is usually attributed to the object rather than to the design. This note reports a preregistered run on the author's own system that returned INCONCLUSIVE, and then reports why it could not have returned anything else. The protocol was frozen and digested before the system under test was present in the execution environment. All six blocking gates passed, including an ablation gate that reproduced the certified reference fixture bit for bit. The mechanism functioned as specified and did not move the metric. With five paired subjects the 95% half-width was 0.1962 against a preregistered margin of 0.15: the design was below its own resolution threshold before execution, and this had not been calculated when the margin was fixed. The stopping rule was respected and no subjects were added after the result was seen. A separate finding outlives the experiment: the defect the work was designed to repair had been described by a superseded metric authority, and the value anchoring that description was later shown to be largely attributable to a scripted driver unrelated to the mechanism it was credited to. Two errata were issued against this record, both against the author's own position.