New benchmark exposes how badly AI struggles with real knowledge work
New benchmark reveals AI's significant limitations in real-world knowledge tasks, with top models solving only 3% of challenges. This exposes a critical gap between current AI capabilities and the demands of practical knowledge work.