Splitting Instruct and Thinking makes sense, the use cases are genuinely different. Curious if this means they'll optimize each path independently or if there's still a shared base.

→ View original post on X — @whats_ai