This is not a relationship between an “old” and a “new” version

This work belongs to Perceptual and cognitive safety . AGI-COS has no single “latest version” covering every task. Different problems, stages, phases, and task-process states call for different professional tracks. A higher version number does not automatically invalidate capabilities that remain effective within another track.

Current applicability

Currently applicable across stages. Perception audits, data provenance, the Unknown Zone, reasons for adopting low-confidence inputs, and pre-deployment preparation should enter uses with major consequences.

Current boundaries

Fixed human–AI decision weights no longer apply; risk metrics cannot automatically allocate governing powers. The cognitive-load and adversarial-resilience elements of 1.2 may be combined as a controlled subtrack.

Continue from this work

Versions may be combined, but permissions must not accumulate. Republication does not restore the current force of historical parameters, governance assumptions, or automatic triggers in the original texts. Conflicting mechanisms are isolated for reconstruction following the continuity audit.

Translation scope

This page translates the public web description. The reviewed English manuscript is available in the English PDF section below; its source edition is identified there.

English PDF

Download the English PDF · V1.1