51–52 of 52 posts
Earlier quoted context omitted.
> And incredibly good ways to assess & test it's weights What weights are you referring to? How does [Claude?] code do that
Look into RLVR (Reinforcement Learning with Verifiable Rewards). It happens during model post-training.
Earlier quoted context omitted.
> And incredibly good ways to assess & test it's weights What weights are you referring to? How does [Claude?] code do that
The hidden virtual weights in reality. Which are often complex & multi-faceted, measuring the rest of reality's weighing, to make broad judgement with. Reality's normal context window is a google deep for even the most everyday of circumstance. The weights exist there, but amid too broad a reality with too many factors for that exacting a use, and are too complected to measure out individually easily. Code is simple…