@TheLifeOutline - old vs L1 vs format-lock

Gates frozen (300k/30d + 150k/15d). Only the GPT query borders differ.

Finding: the "ancient humans drinking alcohol" drift you flagged is ALREADY in the OLD prompt (score 7, shown). Widening to L1 actually pushes that drift DOWN below the visible top-8 while finding MORE real same-format clones.

1 - Old (current production)

Tight prompt →

Finds Lumicus, Master POV, Wallace - but ALSO renders "When Did Ancient Humans Start Drinking Alcohol?" at score 7. Same drift, fewer fresh clones.

2 - L1 Mild (my pick)

Mild widening →

Most real format-clones: Hypothetically (POV 2526), Master POV (Lottery), Wallace (How Every Member of The Seven Died), Dark Ledger (Every Level of Wealth). The ancient-humans drift drops to score 4 - off the visible top-8.

3 - L1 + format-lock

Mild + keep the title format →

Purest top (all "your life as every level/rank"): Lumicus, Master POV, POV Pulse (MMA), Wallace. But fewer finds (only 6 cards) and still lets minor score-3 drift in the tail.

Art Of YT - query-width verdict - dev code, gates frozen