CivicTwin - Proof-Carrying Rule Twin

CivicTwin helps multilingual, first-time micro-entrepreneurs answer a question ordinary search and chat systems cannot: what exactly changes if I change one operating choice? The initial experience follows Sofia Reyes, an Oakland mobile-food founder. Her vehicle type, location, health status, and vending duration activate requirements across city, county, and state sources. CivicTwin converts her spoken intent into structured operating facts, compiles official evidence into typed rules, evaluates those rules with a deterministic constraint solver, and emits an inspectable dependency graph. The winning demo changes Sofia's vending duration from 60 minutes to 30 minutes. The model recompiles the graph and removes the restroom-agreement dependency. When the changed pushcart scenario lacks sufficient fee evidence, the answer remains UNKNOWN and an evidence gap is flagged instead of hallucinating a zero-dollar fee. Every decision opens a proof receipt containing the triggering fact, rule identifier, source URL and clause, retrieval date, confidence, and unresolved boundary. The product was built primarily in native.builder: its agents created and refined the React/TypeScript app structure, multi-route UX, stateful intake, counterfactual controls, trust surface, proof-receipt interface, and public deployment. Partner-ready adapters connect Speechmatics for multilingual transcription, Bright Data for official-source acquisition, and the AI/ML API for rule compilation and counterfactual auditing. Deterministic demo fallbacks are visibly labeled when credentials are not connected. CivicTwin is not a chatbot or a static report. It is a proof-carrying regulatory world model. The business wedge is Oakland mobile vending; the expansion unit is a versioned jurisdiction pack for founders, SBDCs, lenders, and city permitting teams. The goal is simple: turn fragmented rules into the next valid move, without hiding uncertainty.