Opus 4.7/4.8/5 have a built-in nitpicking as anti-syphcophancy. Fable too. The models are structurally incapable of not nipicking. I suspect the nitpicking hel

Opus 4.7/4.8/5 have a built-in nitpicking as anti-syphcophancy. Fable too. The models are structurally incapable of not nipicking. I suspect the nitpicking helps engagement metrics and it doesn't harm RLVF outcomes, so there's just not good signal against it.