This paper describes a file-based orchestration architecture for context-window-spanning AI-assisted knowledge production at book length. It combines quality gates, lane separation, promotion contracts, and deterministic session rehydration into a real deployed system documented over 233 production runs on a single long-form project. The paper is a descriptive, experience-grounded systems report rather than a controlled comparison study.
Brett Westerlund (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: