AI21's Yoav Goldberg calls for strong 'open-thought' models with fully transparent CoT

ChrSzegedy · x · 2026-11-21

AI21 co-founder Yoav Goldberg argues that as scientific progress increasingly relies on AI, there will be an acute scientific need for fully transparent models — especially at the chain-of-thought level.

He predicts big labs will resist this for fear of others training on their CoTs, which is why the community needs strong new "open-thought" models as soon as possible.

Original post →

More from AGI Musings

AGI Musings channel →