The quarry-budget candidate's ordinary seed-3 two-policy game has a staffed metal/weapon economy and14soldiers, but is still unfinished100,000. At the next normal scouting decision, tick100,800, the public AI emits no order even though513tiles remain unexplored and no enemy military building is visible.
All six exact rays through the home building are already explored to their map edges. The current scout policy searches only those rays, so it never visits the remaining off-axis areas. This is a separate military-planner defect, not insufficient weapons or an engine command refusal. The captured checksum is b0ec2e9a, and conservation is clean.
Owning source: economicAI military scouting. Preserve the existing ordinary nearby/ray behavior where useful, then explore reachable known frontier when rays provide no new destination. No enemy-location injection or map revelation.
Acceptance:
- Retain the ordinary captured no-order checkpoint, source hash and provenance after the quarry-budget candidate is published.
- Reproduce the failure through the public observation/economicAI seam at the normal scouting cadence, with14troops and513unknown tiles.
- Select useful known, walkable and reachable frontier outside the six rays, respecting the player's actual knowledge and movable-garrison guard rules.
- Use ordinary commands to reveal new land and demonstrate deterministic mid-scout restore, clean conservation and no guaranteed command refusals. Do not grant visibility or use concealed enemy positions.
- Cover fully explored maps, unreachable frontier, sole guards and an existing useful ray destination.
- Repeat normal matches and full source/browser/exact-main verification; keep broader military/religious AI and original Gold behavior claims open.
Related #74, #76, #198 and #4. Engineering policy against the current Roman profile. The generated checkpoint contains no original raster/audio.