Three of the twelve laws share a grammatical shape. Law I says capability does not imply authority. Law II says the absence of observed failure is not renewal. Law XI says success is not a licence to expand. Each names something true about a system and denies that it settles something about the system’s authority. Read separately, they look like three cautions. Read together, they are one principle applied three times, and the principle is worth stating on its own because the method depends on it more than on any single law.
The default inference
Institutions convert facts into permissions constantly, and mostly without noticing. A tool becomes reachable, so it is used. A model gets better, so its scope widens. A system has run for a year without visible incident, so the question of whether it should keep running stops being asked. A pilot goes well, so the second site is a rollout rather than a second decision. In each case a property of the system, one that was measured or observed, is treated as if it were a grant, which is something an institution makes.
The conversion is not stupid. Capability is evidence that a wider grant might be justified. A clean year is evidence that the original grant was sound. Success is evidence that expansion could go well. The error is not in treating these as evidence. It is in treating them as the decision, so that the grant widens or renews with no moment at which anyone was responsible for widening or renewing it.
That missing moment is the whole problem. Authority that grew without a decision has no record of what justified it, no scope that anyone chose, and no one whose name is on it. When it later has to be narrowed, there is nothing to narrow it back to.
Why the refusal has to be structural
A rule that said “do not confuse capability with authority” would be true and useless, because nobody confuses them on purpose. The conversion happens in the gap between two systems of record: the one that tracks what a system can do and the one that tracks what it may do. Where those are the same system, or where the second does not exist, the conversion is automatic. The permission surface is whatever the capability surface happens to be.
So the method does not state the refusal as a rule of conduct. It states it as a boundary between objects. Capability lives in a catalog, discovered and tested. Authority lives in a grant, with a holder, a scope, a basis, and an expiry. The catalog cannot write to the grant. A capability that appears in the catalog and not in the grant is reachable and refused, and the refusal is a check a harness can run rather than a norm a reviewer has to remember.
The same construction carries the other two laws. Renewal is a transition on the grant with a recorded basis, and Law II’s clause holds that a grant with no revocation conditions is a transfer rather than a delegation, so a grant cannot renew by the passage of time. Expansion is a transition on the grant with its own evidence and correction-capacity check, so a grant cannot widen by accretion. In each case the fact that tempts the conversion is still recorded, as evidence, in the place evidence belongs. What it cannot do is cross into the authority object on its own.
The family
Once the shape is visible, the three laws are the instances that matter most for delegated intelligence, not the only ones. The dependence essays on this site describe a fourth conversion, in which an institution’s reliance on a population’s corrective labor is treated as a fact about staffing rather than a claim the population has on the system, and a fifth, in which a population’s reliance on the institution is treated as licence to direct it rather than as a duty owed to it. The essay on absorption describes a sixth, in which a metric that improved because someone caught the failure is read as a metric that improved. These are not the same mistake dressed up. They are different facts refusing to become different permissions. What they share is the direction of the refusal: from the observed to the authorized, nothing passes without a decision.
A reader may reasonably ask whether this is a principle or a habit of suspicion. The test is whether it says when conversion is permitted, and it does. Conversion is permitted through an explicit state transition, recorded, with the fact that motivated it attached as its basis. That is a positive procedure, and it is what the authority grant and its state history exist to carry. The principle does not forbid an institution from widening a delegation because the system got better. It forbids the widening from having happened without anyone doing it.
What this is not
It is not a claim that authority should be hard to get, or that grants should be narrow. An institution that trusts a system may grant it a great deal, and the method has nothing to say against that. The principle governs the route, not the destination. A wide grant made on purpose, with its basis written down, is exactly what the method is for. A wide grant that accumulated is what it exists to prevent.
It is also not a claim that facts about a system are unreliable. Capability measurements, incident rates, and pilot results are the best evidence an institution has. The point is that evidence is an input to a decision and not a substitute for one, and that the difference is only real where the two are held in different objects with a recorded transition between them.
What the laws take from this
The method does not ask a reader to accept that there is a family here, or that the instances beyond the first three are correctly described. It takes three constraints and binds each to a clause.
From Law I it takes that no increase in capability increases authorized agency, and that authority expands only through an explicit state transition. From Law II it takes that the burden of proof rises with the duration and consequence of a delegation, and that the absence of observed failure is not renewal. From Law XI it takes that expansion is not the default reward for success but a new authorization decision.
An engineer who rejects the essay’s reading can still adopt all three on narrower grounds: a grant that has no recorded basis cannot be audited, cannot be narrowed to anything, and cannot be defended when it is challenged. The theory says why the conversions are refused. The method only insists that they be.