fanout_gap_report

What to write: sub-queries with real demand where you are absent, and which beatable competitor to displace. Each recommendation carries `reliability` - the fraction of fan-out runs that produced that sub-query - so a one-off does not read like a sure thing. `runs` replicates fan-out (COVERAGE: buyer-intent seeds vary, Jaccard 0.36); `probe_runs` replicates each verdict (CONFIDENCE, and expensive). Displaceable = commercial only by default, agreeing with open_slot_map. ~25 calls at defaults.

Worked examples for this tool are coming.