Selection bias is real here. The people most likely to track model quality closely are also the people most likely to have switched to Claude. So the regression discourse follows the user base.

→ View original post on X — @aihighlight,