Structured Truth-Finding with AI-Augmented Teams: A Multi-Node Epistemic Protocol Preserving the Scientific Method Under Capacity Scaling

This paper describes a working protocol for AI-augmented research developed during an extended investigation in theoretical physics. The protocol specifies roles and actions within the investigation, an epistemic labeling discipline for tracking confidence levels, a session architecture with explicit handoff documents, multi-node adjudication procedures when AI verifiers disagree, and registries for dead ends and parked ideas. The protocol is presented as a practice-derived pattern rather than as a validated methodology: we can describe what has worked in one investigation and name the failure modes we have encountered, but we cannot demonstrate the protocol produces better results than alternatives. The paper is offered as a starting point for other practitioners working on the broader question of how AI participants should be used in serious investigative work. This paper describes a protocol developed during work on Möbius Field Theory (MFT), a speculative unified physics framework. The protocol's utility does not depend on MFT being correct; the two questions are independent. Readers interested in the physics should consult the MFT preprints directly.