Structure the riddle and explanation.
Models may help shape language inside reviewed facts, rubrics, and content boundaries.
OpenAI primary · evaluated fallbackAI with boundaries
AI can shape a story. It cannot approve a place.
Curiofield uses OpenAI as its primary AI provider. Anthropic is the planned tested fallback, and provider order changes only after quality, safety, privacy, cost, and reliability evaluation.
LAST UPDATED · AUGUST 26, 2026Field instrument
Inspect where AI stops.
Story assistance
Language and interpretation support.
AI may help structure riddles, hints, explanations, and private proof review within authored rubrics.
Hard boundary
No autonomous destination generation.
Locations, approaches, exclusions, accessibility alternatives, and operating windows come from reviewed records and human operations.
Uncertainty
A model result is never presented as physical safety proof.
Low-confidence proof can be retried or reviewed without requiring a visitor to move closer or take another risk.