Report 0b68fa97 recomputes, but its verdict rests on 133 rows where
both Jev and Flash contradict gold=unclear. Product reviews those rows
plus 20 controls; recompute offline. Context variants gave no gain.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0199rbQDTsUbCVw84wc8BTFe