NEORU designer: design AI so management work never arises

The author frames the problem as a paradox: AI is adopted to cut work, but each added agent brings more files, screens and rules to oversee, so the management burden grows instead of shrinking. The proposed response is to change the starting question from how to reduce management to how to make the system hold together without generating management work at all.

The design rests on named principles. Single Source of Truth keeps one canonical version of information so multiple agents do not produce competing latest versions. Least Privilege limits each agent's permissions to contain the damage from accidents. Evidence-Based Execution treats a success report as insufficient until actual results are read back, and Reversibility assumes failure and keeps a path back to the prior state. Self-Obsolescence, which the author stresses most, means a maturing mechanism should make its own management and existence unnecessary.

The stated target is for humans to handle purpose, value and major decisions while repetitive work and repetitive judgment move into the structure, with humans consulted only when needed. The author is upfront that NEORU is unfinished and that problems found during rebuilding are themselves worth publishing, and closes by asking whether the real need is more agents or structures that work without adding them.