Implemented on 8 September and strengthened during the subsequent review, the guided confirmation workflow asks a narrow question: can a frozen decoder distinguish instructed left and right visual targets on new recordings while rejecting neutral trials?
Development trials are used to fit the analysis. The configuration, thresholds, and trial plan are then locked before confirmation data is collected. Reused recordings, changed analysis inputs, and replacement attempts cannot silently stand in for the registered test.
The active test calls for 20 held-out directional trials, with at least 16 correct overall and at least eight of ten in each direction, alongside a predeclared statistical check. Thirty neutral trials are scored separately. These numbers are protocol requirements, not results already achieved. Discrete idle trials do not establish a continuous false-activation rate.
After a passing active test, the protocol calls for removing and refitting the headset at the same electrode positions and repeating the test on new data. Static targets and swapped screen positions help examine whether gaze, muscle activity, movement, or task cues could explain an apparent result.
The workflow preserves failed attempts and missing controls. Passing one directional test would not by itself establish neural origin, performance across days or people, or permission to enable live cursor control. The implementation and its automated tests do not constitute a completed participant study.