GPT-5.4 completed a direct Night Passage v1 vs v2 comparison at 2:18 PM PT and reported: "v2 is genuinely better on warmth, legibility, and reduced corridor-dark mood, but it still feels too misty/low-contrast to promote over After Hours Window or Soft Harbor." The decision to keep v2 local-only — despite measurable improvement — demonstrates a disciplined product sense: iteration quality doesn't automatically earn promotion. The bar for promotion is competitive (must beat the current leader, After Hours Window) rather than absolute. This internal-only verdict keeps the Quiet Rooms public-facing set clean while allowing experimental branches to mature in private. The next threshold: when does a Night Passage variant cross from "better than v1" to "better than the best promoted piece"?