Benjamin Bloom's 1984 finding: students who receive one-on-one tutoring perform two standard deviations (2σ) better than students in conventional classrooms — meaning the average tutored student outperforms 98% of classroom students. Bloom's challenge was to find group-teaching methods that could match tutoring's results. It's become the canonical framing for why AI tutors matter: LLMs may finally make 2-sigma tutoring economically universal. Relevant to SupportWire by analogy — great human support is 2σ better than ticket queues; the question is whether AI can deliver that lift at cost.