Every change to how Votr scores, recorded in public. Entries describe what changed and what it means for existing scores; the engine’s internals stay behind the published boundary.
When you rank your tags from S to D, that ranking now shapes your match: an issue you hold strongly — by one strong choice or several that agree — counts for more than one you only lean on, and an issue where your own choices pull both ways counts for less. Before this, how strongly you felt only mattered where your choices disagreed. Your profile page places you on each issue by the same rule. This is a major change: some alignment numbers moved by several points. What the politician’s record shows, and how it is read, did not change.
Every position derived from a vote is now weighted by how old the vote is — a smooth half-life rather than the previous stepped schedule — and no vote ever drops out: past a floor, older votes settle into a politician’s archival base. The rule is the same for federal and state records and is applied when a score is read, so existing scores re-weight in place without being recomputed or deleted. The half-life length and the floor are part of the sealed engine.
Every bill page now shows where the bill sits in the dimensional model — which dimensions it touches, which direction it moves, and the cited passage from the bill text behind each placement. Nothing about existing scores changed; the receipts that always backed them became visible.
New scores are stamped v8. The switch was made only after v8 and v7 produced identical results on the served comparison set, so no one’s alignment number changed on cutover day. v7-stamped scores remain readable and comparable within their version.
The fixed catalog of political dimensions was replaced by a governed registry: each dimension carries a ratified definition with cited anchor sources, new dimensions enter through a discover-curate-ratify process, and every stored score is stamped with the methodology version that produced it. Scores are never silently re-interpreted when the model evolves.
We adopted our current publication posture: the scoring engine is a trade secret while Votr operates (so it can’t be gamed), and its full source is held in escrow for automatic public release under AGPL v3 if Votr is ever captured away from its mission or shuts down. The public trust surface is this methodology description plus per-score receipts.
Alignment computed on a fixed dimensional substrate, refined across seven internal versions. v7-stamped scores remain retained and comparable within v7.
Older internal iterations (v1–v6) predate public scoring and are covered by the escrow: if it ever releases, the full lineage releases with it.