Three actionable fixes proposed. (1) HIGH: Add minimum drama score threshold (0.15) to prevent low-value “deepening_bond” encounters from wasting 8-12 LLM calls each — encounters with drama=0.03 shouldn’t trigger. (2) HIGH: Remove duplicate COMMUTING_TO_WORK pathfinding at pet-brain.ts:1890-1909 that redundantly calls getNearestPOIByType + steerAlongPath when the generic task-commute handler at line 2182 already covers it. (3) MEDIUM: Increase commuting fast-path scan interval from every 4 to every 12 ticks, cutting expensive spatial scan frequency for commuters by 3x to address tick overruns reaching 500-600ms.
Changes
- (HIGH): No minimum drama score for encounter initiation — low-value encounters waste LLM budget
- (HIGH): Duplicate `steerAlongPath` + POI lookup for COMMUTING_TO_WORK wastes computation
- (MEDIUM): Commuting creatures exit fast path too frequently on scan ticks
- (LOW): LLM phrase echo chamber — “world is holding its breath”