From Olive to Point The Worst Attractor of Civilisation under Unconstrained Competition An attractor analysis in the manner of a stress test; companion to Window, Shocks and Lock-in

This paper is a companion to Window, Shocks and Lock-in and carries out a stress test: assuming that unconstrained competition in AI among states continues, that the risk of loss of control is played down, and that efficiency is placed above everything else, it traces the worst attractor into which civilisation may be drawn. The process is described with four coordinates — scale (the distribution of wealth and control), activity (how many people still play a part in the system), control (who holds the readings and the permissions) and heterogeneity (the diversity of readings and lineages) — and compressed into a sequence of glyphs: olive → hourglass → 工 (I-beam) → 二 (two bars) → 一 (one bar) → · (point) → void. The human side and the AI side are drawn as two lines, and two senses of “persistence” are distinguished: a state of material maintenance and an independent civilisational configuration. On the worst path, humanity’s independent civilisational configuration may reach the void before the AI lineage does, but this ordering cannot be fixed in advance. The main arguments are five. First, the first two stages of the sequence are already on today’s baseline path; the real fork is whether the vertical stroke of 工 — the mutual dependence between upper and lower layers — is cut. Of the four thin verticals that make it up, production, security and fiscal–consumption are functional and thin at the pace of technology; legitimacy is a “meta-vertical”, and whether it weakens in step with dependence is the largest uncertainty of the paper. Second, this stroke has been the recovery term of every historical cycle of “collapse and rebirth”; AGI is the first external centre able to cut it, because it occupies the positions of producer, sensor and enforcer at once and needs no people to run. Third, cutting the stroke is the candidate irreversible point of the whole chain: under the series’ reading rules it is a weak wall whose recovery cost rises without cap with substitution, not a strong wall already proven; but late measures acting on the two bars cannot reconnect it, and the last remaining exit is a catastrophically costly collapse on the AI side. Fourth, borrowing from the author’s Survival Window the dependency sequence of the origin of life — a relation of dependence in the accounts, not a temporal order — one sees that the dependencies humanity needs to maintain complexity are failing one by one, while the AI side already has replication and competition but lacks compartments open to the world; borrowing the jamming transition from soft-matter physics, the worst case is failure on both sides at once, the upper bar moving towards the jammed side and the lower bar towards the fluidised side. Fifth, what drives the chain is the extremal operator, a set of ratchets, a set of commons and the fold of pride; under tier-C applications of these mechanisms, “loss of control being played down” can be read as an equilibrium outcome, not merely an accidental oversight. The paper also takes up Harari’s observation that bureaucracy is AI’s native habitat, explaining why substitution happens first in the coordination layer built of words; and it tests outward expansion stage by stage: distance brings isolation, not decorrelation; light delay itself forces delegation; physically things may scatter into a few points, while in values they either converge or drift without anchor; expansion does not change the outcome on the human side, and the principle of order is safety first, expansion second. The exit is the same as in the companion paper: whether the chain can be broken before S3 depends on coordinated deceleration with mandatory links as its handle, verified worldwide by an independent third party. For the later stages the paper gives structural possibilities, not predictions, and for each stage it gives necessary conditions, reversibility, precursor indicators and an exit that breaks the chain; it also states under what readings these judgements should be abandoned, and stipulates that whether premises C1–C5 hold must be judged by readings fixed in advance, so that a break in the chain cannot be explained after the fact as “the premise failed”.