What is affected
Bills can carry many policy tags, legal effects, affected populations, local-government layers, and agencies. A tag can be explicitly supported or an inference, and we show that distinction.
Signals and uncertainty are displayed separately from verified legislative actions.
Bills can carry many policy tags, legal effects, affected populations, local-government layers, and agencies. A tag can be explicitly supported or an inference, and we show that distinction.
Author influence is based on independently verified roles: chamber leadership, committee chairmanship, relevant committee seats, and legislative tenure. We do not assign prominence just for party or media attention.
The atlas shows the strongest hearing signal supported by the record. A referral alone does not guarantee a hearing.
Passage requires a separate model: committee action, chamber calendars and votes, companion progress, cross-chamber support, deadlines, and executive position. No passage probabilities are currently claimed.
Each action, bill version, fiscal note, referral, hearing notice, companion filing, and author role can add or change signals. The dataset preserves source timestamps and links to the state record.
Luna is the AI model that reads each bill’s full text. It writes the “In brief” line and the summary, adds topic, agency and search tags, and marks the passages its notes rest on. Everything it writes carries an AI label, can be wrong, and is never official: the bill text and the legislature’s record are. Review progress shows which bills it has read.
Automated agency extraction from captions is incomplete. Predictions in early prefiling are particularly unstable, and committee assignments after the election must be verified before influence scoring.