Hacker News Reader

Pretty sure those are just summarized traces, which obscure the valuable reasoning traces used by the actual model.

yeah i know they are summarized/condensed, but they are still an order of magnitude more useful that what they competitors present to users.