Post-training has a ceiling too: AI will match, not surpass, the best humans

Liu_eroteme · x · 2026-09-29

Responding to a discussion on pretraining limits, one commenter argues post-training is also capped: it can only approach the best policies external verifiers can verify.

His prediction: we'll get models "equal to or slightly better than the best humans at any task," but not genuinely superhuman intelligence — a notable contrarian take in the scaling debate.

Related event: RLVR Is Capped by Verifiers, Limiting Path to Superhuman AI(3 posts)→

Original post →

More from AGI Musings

AGI Musings channel →