AI language models' context windows have grown 3,906x since 2017 while human attention spans have shrunk to a fraction of their capacity, creating a widening cognitive gap.

LLM context windows exploded from 512 tokens in 2017 to 2 million tokens by 2026, doubling every ~14 months

Human Effective Context Span declined from 16,000 tokens (2004) to just 1,800 tokens by 2026, based on reading-rate analysis and behavioral data

The AI-to-human cognitive ratio shifted from near parity at ChatGPT's launch (November 2022) to extreme divergence by 2026

This self-reinforcing 'Cognitive Divergence' reflects how delegation to AI may further erode human sustained attention capacity