A local-first autonomous intelligence layer for OpenRCT2 that founds, governs and remembers an amusement park across decades.
Mission profile
- Operate AFK under deterministic budgets, permissions, snapshots and rollback.
- Build historically coherent parks rather than cosmetic period skins.
- Model each visitor as a persistent person with memory, relationships and beliefs.
- Learn from industrial failures, bankruptcies, accidents and strategic mistakes.
- Generate and validate rides through search, simulation and engineering constraints.
Architecture
A park as an institution
Helter Skelter treats the park as a governed organization with a constitution, treasury, engineering doctrine, safety vetoes, institutional personality and long-term memory. Spectacle is subordinate to survival and coherence.
Artificial personhood without theatre
Visitors receive persistent autobiographical and social models, but the project makes no mystical claim of phenomenal consciousness. Population behavior remains hierarchical: deterministic needs at scale, group dynamics above them, and richer cognition for salient individuals.
Local-first by design
The useful system must run on a Linux workstation with an RTX 5080 and scale upward to RTX 5090 or RTX PRO hardware. Cloud services may accelerate research, but they must never become mandatory for core operation.