Part III

A permission lasts while the grantor remains willing. A right begins where that willingness ends.

This is the kind master problem. A platform may behave generously, a custodian may honor every withdrawal, and an executive may exercise extraordinary restraint. None of those facts changes the position of the person whose liberty remains subject to a power the person cannot bind, contest, or replace. Republican freedom begins with that latent relation: the power already exercised matters, and so does the power that remains available at another's pleasure.

Non-domination therefore asks for more than favorable outcomes. It asks whether a claimed liberty survives management change, insolvency, a revised risk model, political pressure, or the moment when compliance becomes expensive. A right must remain usable at the adverse moment, when the institution bound by it would prefer not to honor it. If the only remedy is an appeal to the same will that withdrew the permission, kindness has been concealing rather than curing the constitutional defect.

Cryptography can alter that relation inside bounded digital domains. A key can remove an intermediary's authorization switch. A commitment can make one form of defection infeasible under declared assumptions. A proof can establish that a specified computation or relation holds without requiring the verifier to trust the party that produced it. These are real constraints, and they matter precisely because they do not depend upon the constrained party's virtue.

Their domain is also exact. A signature does not prove that its human controller acted freely, a proof does not establish that the encoded rule was legitimate, and self-custody does not protect the body holding the key from coercion, fraud, device compromise, censorship, or exclusion from the institutions that recognize the asset. The key changes one relation. It does not make its holder sovereign over the surrounding world.

The same boundary applies to verification. A right to inspect code is formal where the affected person lacks the artifacts, expertise, time, or forum needed to make the inspection consequential. Constitutional access does not require every citizen to operate validator infrastructure or become a cryptographer. It requires a usable independent check, whether direct, representative, or public, completed before the appeal expires and connected to an authority capable of changing the result.

Even then, proof settles only what the proof contains. It may establish execution, assist with provenance, and expose a mismatch between a declared rule and an output. It does not by itself establish that the inputs were lawfully obtained, that the inference is valid for this person, or that the institution may use the inference for this purpose. Part III therefore treats cryptographic constraint as one layer of freedom rather than its total architecture. Law, independent review, portability, and public reason remain necessary where the key returns to the body and the computation enters a life.

History gives no reason to expect these tools to distribute power by themselves. Open protocols can become the substrate of concentrated platforms, self-custody can be routed through custodians, and nominally transparent systems can depend on interfaces and institutions the governed cannot replace. Capture is not answered by the word decentralized. It is answered, imperfectly, by making dependencies visible, powers contestable, and rights practically available to those without the resources of the operator.

Part III asks what freedom requires once good intentions have been removed from the premises. It begins with the master, follows the key to the edge of its domain, and ends with the means by which a person can test the proof used against them. The result would leave power in place while depriving some powers of the ability to act silently or finally merely because they own the interface.

The remaining task is institutional. Rights that survive adverse will still require charters, mechanisms, interpreters, emergency rules, and forums among overlapping jurisdictions. A constraint must run, and the case it cannot settle must have somewhere legitimate to go.

Part IV begins with the rule and its remainder.

---

A right is not a promise of good behavior. It is a position that remains usable when good behavior ends.