AI Delivers Two-Thirds of Cognitive Task Results, Human Judgment Holds Final Value

Pseudonymous account @signulll posted a principle describing how current AI models reach roughly two-thirds of a good result on most cognitive tasks. The remaining portion requires taste, judgment, iteration, domain expertise, and originality, which together hold nearly all the value. Danielle Fong, a physicist and founder, amplified the post. The observation notes that ongoing model improvements continue to leave that final segment dependent on distinctly human contributions rather than automation alone.