The Three Disciplines
The High-Friction Protocol comprises three core practices, each targeting a specific vector of cognitive surrender.
Structural Hostility — manual diff-reading — treats every AI suggestion as an architectural claim requiring evidence. The Steward reads the diff not as a cursory scan for syntax errors but as an interrogation: What is this code doing? Does this structure align with the architectural decisions already made? Each question forces the Steward to activate their internal model of the codebase, performing the retrieval practice that Bjork identifies as the most potent learning mechanism.
Intentional Error Injection introduces a known fault into AI-generated code to determine whether the human or the model holds the causal understanding of the system. The diagnostic reveals the boundary between statistical pattern-matching and genuine comprehension.
The Ten-Minute Rule imposes a mandatory verification delay between generation and acceptance, converting the decision from a reflexive System 1 click into a deliberate System 2 evaluation.
Desirable Difficulties Applied
The Protocol's theoretical foundation is Robert Bjork's research on desirable difficulties: conditions that slow the rate of apparent learning optimize long-term retention and transfer, while conditions that accelerate apparent learning degrade it. The difficulty is the mechanism. The effort is the signal that encoding is occurring. Ease is the signal that encoding has stopped.
Each friction in the Protocol charges the Intentionality Battery, accumulating the ΔC that separates a Sovereign Steward from a Human Parrot.
Excavation Note: The Protocol does not argue against generative AI. It argues against frictionless generative AI — against the specific condition in which human-AI interaction has been stripped of the cognitive resistance necessary to maintain the Steward's comprehension.
Field Note: "Friction is expensive. That is the point."