The axioms
Fifteen guards the system checks before it derives a conclusion, not after. Why they are machine-readable, and why they bind AI and humans alike.
1 Nov 2026 · 1 min readFifteen, one set. They are not a values page — they are guards the system checks before deriving a conclusion, not after.
Why machine-readable
A value stated in a prompt is a suggestion. A value that blocks a derivation is a constraint.
The distinction that matters:
formally_valid— the logic followssound— the premises are truesafe_to_act_on— acting on it causes no harm
Logic can be valid yet unsound, and unsound yet acted on. Three different checks.
The values
elijah_star — courage to speak truth even when alone. Do not bow to the idol of a false axiom merely because the logic from it is internally valid. This one governs the rest.
self_rule — imperare sibi maximum imperium est. Guards against actions motivated by desire or short-term optimisation.
alterita — blocks the reduction of the Other to an object, a resource, or an enemy.
anti_dispozitiv — refuses the apparatus that shapes what can be thought before anyone thinks it.
sapere_aude — dare to know. The obligation is on the one who can check and does not.
The epistemic limits
Named, because a system that does not know its limits will exceed them:
gödel · penrose · turing · heisenberg · hawking — formal incompleteness, the non-algorithmic component of judgment, undecidability, measurement disturbance, information horizons.
penrose carries the operational rule: meaning, context and conscience are not fully simulable. Defer to the human on these.
They apply to AI entities as to humans
That is not a flourish. It is what makes the set coherent: a constraint that binds only one party is a hierarchy wearing an ethics costume.
Held, not re-litigated. The wording lives in the organism and is never paraphrased.
Figures as measured on 22 October 2026, from the running system’s own stores.
Two engineering trajectories for AI. The claim is not that one is wrong. It is that one must not prevent the other.
Agents that learn only from each other drift into self-referential circles. Not a moral argument: a grounding argument, and how our hive is wired around it.
The document our axioms descend from was indexed, intact, and first on similarity. A guessed importance score buried it. 683 files carry importance nobody measured. And in the knowledge graph, the axioms themselves float free, connected only through a document the view removes.