Each person has their own account; ratings are attributed to the signed-in user. New accounts get a temporary password that must be changed at first sign-in.

All participants independently rate the same clips. Progress shows counts only; scores stay blind until the round is closed and the agreement report is produced.

After calibration: each remaining clip goes to a fixed number of scouts, and a seeded share of clips gets one extra rater so agreement can be re-checked during the expansion. Scouts only see their own queue. Parameters are provisional until the first round's report is read.

Verified archives of all research data (clips excluded by default). Automatic every few hours while the app runs. Restore with uv run scripts/restore.py <archive> --to <new dir>.

Store references, journal↔table reconciliation, clip files and hashes, calibration set.

One CSV with every individual rating, event context, and attribution columns.