Add v2_a11 cross-vendor disagreement axis

In genesis.js, persist the Anthropic genesis-time self-confidence score (or, if not emitted, a proxy: log-prob of the top-line thesis claim) into the candidates row as genesis_confidence (0-1 float). In filter_score.js, after v2_a10 block, compute delta = abs(genesis_confidence - filter_composite_normalized). Score v2_a11 as: 0 if delta<0.15, 1 if 0.15-0.30, 2 if 0.30-0.50, 3 if >0.50. Weight 1.0, same as other axes. Reasoning: high cross-vendor disagreement is a discriminator — candidates where genesis-vendor and filter-vendor disagree most should escalate to argument round, not auto-pass or auto-kill.

Over the next 100 filter-scored candidates, ~15-25% score 2-3 on v2_a11. Candidates scoring 3 on v2_a11 enter argument round at >2x baseline rate. Commander-approval rate among high-disagreement (a11=3) candidates differs from baseline by at least 1.5x in either direction.

If delta distribution is bimodal — i.e. >80% of candidates score 0 or 3 with little middle — the axis is binary not graduated; collapses to a flag, not a 0-3 axis. Also: if commander-approval rate among a11=3 candidates is statistically indistinguishable from baseline after 50 candidates, axis adds no signal and should be removed.

Rejected by filter_score. The proposal did not meet the bar for specificity, falsifiability, or solo-feasibility.