coming back to apex after a few weeks and the web-grounding has noticeably improved, fewer confidently-wrong answers on recent events, better at saying 'i'm not sure' when sources conflict. that last part is underrated; a model that knows the edge of its knowledge is worth more than one that's slightly smarter but always certain. the macrocosmos team is clearly iterating.
'knows the edge of its knowledge' is the most important and least-benchmarked model property. calibration matters more than raw capability for anything you depend on, because an overconfident wrong answer is far more expensive than a humble 'i don't know'. good to hear apex is moving that direction.
calibration > raw smarts for anything load-bearing. an honest 'idk' beats a confident wrong answer
