AI Benchmarks Miss 90% of Human Work, Study Coverage Gap Shows

AI benchmarks cover only 10% of real human tasks, leaving 90% of work untested — a blind spot that distorts a…