Cognitive amplification is a methodology for human-AI collaborative work in which AI serves as instrument rather than co-author. This paper introduces a two-layer operational framework - an advisory layer that challenges and sharpens human intent, and an implementer layer that executes against specifications the human has already developed - and argues that authorship remains entirely with the human when the architecture is enforced correctly. The framework emerged from 18 months of adversarial practice building the YIM Project, including 50,000+ documented conversation turns across 250+ sessions and the development of the Core Six taxonomy of AI defensive behaviors (Taylor, 2026, doi:10.5281/zenodo.19423182). It is demonstrated across two domains: a sustained independent research program and a real-time legal crisis managed without an attorney. The central claim is that the bottleneck in serious knowledge work is rarely raw intelligence - it is the gap between what a practitioner understands and what they can articulate, organize, and sustain across a complex project. Cognitive amplification addresses that gap without displacing the thinking that belongs to the human.
Ernesto Taylor (2026) studied this question.