Mercor
Edward Hu: The Reason AI Solved 722 Math Theorems But Still Can't Answer Your Pizza Phone
Mercor's head of AI modeling Edward Hu discussed on The Cognitive Revolution podcast how Mercor's Apex Accounting benchmark found Anthropic's Opus 5 scored full credit on all 20 attempts versus roughly 37% for twelve licensed CPAs, while cautioning the test only covers a narrow slice of real accounting work.
View Original