Aggregate: a count or summary across many records (e.g., "42 open streetlight requests in this area") rather than information about one person.
Audit log: an append-only record of administrative and safety events. Rows can be added, never edited or deleted.
Build: the track where residents and volunteers co-design a small tool.
Data steward: the named person responsible for what data is collected, how it's protected, and deleting it on request.
Federation: optional exchange of signed, aggregate-only bundles between neighborhood nodes.
Hallucination: when an AI model states something false with confidence.
Learn: the AI literacy track.
Local model: an AI model running on hardware in the hub rather than on a company's servers.
Node: one neighborhood's installation of the software, usually one SQLite database file.
Ollama: free software for running open AI models locally.
Problem Brief: a one-page description of a local problem: what, who has authority, what evidence, and the smallest next move.
Provenance: where a piece of information came from; every fact should link back to its source.
Red-Light Test: the six questions in the Pilot Rubric that instantly reject a proposal.
Safety driver: core/db.js, the code that refuses banned data structures.
Solve clinic: a session where residents turn real problems into Problem Briefs.
Socrata: the platform behind Kansas City's open data portal.
Legistar: the platform many city councils use for agendas, minutes, and votes.
Toothbrush test: does anyone use this at least weekly? If not, reconsider it.
Trust level: 0-5 scale in the identity system, earned through verification and vouching, used to decide who can vote on what.
Wedge: a single, sharp problem a tool focuses on.