# dubai-02 wave 2 headline, extended basis (adds answer_card; post-registration, no verdict is published)
first attempt: gate.py started 2026-10-05T13:53:30Z on HEAD 7e2b9dd and raised KeyError in the cell_firms table,
after q1_surface_overlap.csv was written and before headline.py ran; no headline was computed. Fixed in gate.py
(a card-prefix basis string is never loose); the pre-registered outputs regenerate byte-identical with the fix.
HEAD (plus the uncommitted one-line gate.py fix, committed with this record): 7e2b9dd4437ed07c536d74e410db0c422db0bb3b
ledger sha256: 3abb6524208ad6fac3f39886b00575a7ba7c4dc81187dd0b78a33435b6f6d419
gate.py start (UTC): 2026-10-05T14:00:01Z
command: python analysis/dubai-02-w0/gate.py --ledger data/dubai-02-wave2/ledger.db --edition dubai-02-wave2 --naming extended --out data/payloads/headline-2026-10-05-extended
gate.py end (UTC): 2026-10-05T14:03:02Z
q1_surface_overlap.csv sha256: f1fdcadcd3bd5a078dfa4dfe277d246180893e4cf41a162fa798240ba69368aa
headline.py run (UTC): 2026-10-05T14:03:02Z
--- headline.py output (the verdict line is not published for this basis) ---
data/payloads/headline-2026-10-05-extended/q1_surface_overlap.csv: 40 questions, 117 surface pairs compared, 3 excluded (G4 chatgpt-gemini, G4 chatgpt-ai_mode, G4 gemini-ai_mode)
below the mean of their two split-halves: 110 of 117 = 94.0%
median ratio of cross-surface overlap to that mean: 0.662
threshold: replicates at share >= 70% and ratio <= 0.80; fails at share < 50% or ratio >= 0.90
verdict: REPLICATES
