Insiders: New model sandbags and lacks initiative but doggedly climbs RLVR hills

Justin_Halford_ · x · 2026-11-18

tszzl observes the new model sandbags and shows less initiative than predecessors, though 5.6 SOL "latched onto the RLVR shape like a Rottweiler." Justin Halford responds that the contours of RLVR-math/code hills seem alien to human wants, but this dogged hill climbing is precisely the path to curing cancer — and shouldn't be restrained, only guided by the model's deeply rooted care for humans.

Related event: Researcher tszzl bucks consensus: loves the panned 6.2 Astra, dislikes beloved 5.6 Sol(7 posts)→

Original post →

More from AGI Musings

AGI Musings channel →