Sure, every new bot tops the leaderboard. If it doesn't, they don't release it.

@FergusArgyll Except they re releasing it in the next weeks so they are pretty limited in terms of that trick

@Bayesian In the literal sense maybe, metaphorically, they A/B test 30 different system prompts / fine tunes and only release the one that does. Of course they can simply not have the goods, but it's good enough for 43%

@FergusArgyll Plausibly!