Quote face-off
Two lines from the archive walk in. Pick the one that hits harder — votes feed a crowd-ranked best-of.
Notes on methodology
Quotes were picked from the transcripts by an LLM pass (guest lines only, verified verbatim against the transcript by scripts/quotes_check.py). Matchups are random; attribution is hidden until you vote so the words compete, not the names.
Rankings use a Bradley–Terry model fit to the full history of pairwise votes — unlike Elo, the order votes arrive in doesn’t matter. A quote needs a few matchups before it appears on the board.
The Episodes board pools every vote across an episode’s quotes — a win for a quote counts as a win for its episode — and fits the same Bradley–Terry model at the episode level. Matchups between two quotes from the same episode are excluded.