How we place candidates and measures
Filled dot: academic score. For anyone who has served in Congress we use DW-NOMINATE (Voteview, UCLA), the standard political-science measure built from every roll-call vote they cast. It reflects their voting record, not their party label.
Hollow dot: AI-estimated. Everyone else is placed by the Jev model (TypeSafe AI), which reads the candidate's own public record (their Wikipedia article and campaign website, linked on each card). It isn't told their party. If those sources don't describe real policy positions, we say so instead of guessing. Measures are placed from the official ballot summary. Confidence is shown on every AI estimate.
Endorsements. Where a candidate has been endorsed by Donald Trump in 2026 we note it, with the news report it comes from (via Wikipedia's maintained list). We're adding other notable endorsers as reliable lists become available.
We never endorse. This is a starting point; read the sources and decide for yourself.