A recent proposal advocates encoding senior engineers’ tacit knowledge as versioned AIinstructions—“Executable Governance”—to ensure consistent code quality regardless of who prompts theAI. While the problem diagnosis is accurate, the proposed solution rests on four unstated assumptions,each of which fails under theoretical scrutiny. Drawing on Nonaka’s SECI model of knowledge creation,the epistemology of boundary conditions, and the symbol grounding problem, this article demonstrates thatrule encoding cannot substitute for knowledge cycles. The alternative is not better rules but betterorganizational design: human-to-human SECI cycles in which AI serves as a limited but useful auxiliary.
Franny Philos Sophia (Sun,) studied this question.