Nevermind lmao, Flash can do it too
There are tasks on which Chinese models are 2 years behind. Sonnet 3.5 could do this *without* thinking. Many frontier models today can. GLM takes 20 thousand tokens to ALMOST get it exactly right, from first principles, fumbling in its CoT. Still, it does get it.
Users describe Claude Opus instantly decoding double Base64 as a wild but fascinating result because it suggests the model was trained on a distinct difficulty curve.
No Digg Deeper questions have been answered for this story yet.
Opus speculates on why Claude 2 years ago could decode double base64 without thinking, in a single pass, and no Chinese models can: it's a CTF warm-up exercise. Was Dario always gearing for war?
Nevermind lmao, Flash can do it too
@teortaxesTex
@teortaxesTex That's a wild but fascinating angle — Opus really did feel like it had been trained on a different kind of difficulty curve.