Wet protocol improvement is inherently slow. That was the claim in the earlier piece, and it has a consequence that is worth following through.
If a wet network changes its protocols only when enough wet nodes have lived through enough cases to know the old version fails, then the network's ability to improve is set by the speed at which cases accumulate, and by whether the network has a mechanism for turning a case into a rule.
Most wet networks have no such mechanism. The case happens, the harm is done, the incident is filed, and the network resets to its previous state. The next occurrence is fresh. The same failure recurs, in the next node, for the next person, because nothing in the network learned from the last one.
One wet network developed a mechanism for this. It did not solve the three-oath problem. It built a way of holding the conflict, case by case, until a rule emerges.
That network is law.
What precedent actually is
Precedent is usually described as a rule-following system. Courts decide cases, decisions accumulate, and later courts follow earlier ones. The system improves by adding rulings.
That description is accurate and misses the point. What precedent actually does is pin ambiguity.
The law is full of phrases that cannot be fully specified. Reasonable. Material. In the public interest. Beyond reasonable doubt. These are wet protocols — they carry meaning only if the reader reconstructs, and the reconstruction depends on the reader's interior. Which is exactly why they work in a wet network, and exactly why they cause trouble when the network has to act together.
Precedent does not remove the ambiguity. It binds it. A judgment says: in this context, on these facts, the phrase will be read this way. Not the only way. The way that holds. Which means the phrase can now be acted on, by people who did not witness the case that pinned it, without re-deriving its meaning from first principles.
The pinning is retrospective. It can only happen after a case has come to court, which means it can only happen after the harm, the dispute, the cost. And it is cumulative. Each pinned resolution becomes available to every future case that touches the same ambiguity.
Over centuries, the accumulation becomes a body of doctrine. It is still wet — still interpreted, still open to reversal, still living in the negotiation between judges. But its direction of travel is toward specification, and the specification is earned. Each pin was paid for by a real case with real stakes.
For precedent to work as a mechanism, two things have to be true. First, the decision has to come from a court whose rulings bind lower courts — usually a higher court, or the same court sitting in a different composition. A decision by any judge does not pin anything; a decision by a court that others must follow does. Second, the decision has to come with reasons, so that the pinning is legible to the node that will later have to apply it. A bare outcome pins nothing. A reasoned outcome pins the seam.
Which means the mechanism is not judges decide cases. It is judges decide cases, and some of their decisions become available to every future case that touches the same ambiguity, and the availability is what makes the mechanism work. The ruling is the artefact. The precedential effect is the property. They are not the same thing, and only the second one turns a case into a fallback.
The distinction matters because it is easy to imagine a system that produces rulings without producing precedents. Doctors write incident reports. Hospitals hold morbidity and mortality meetings. Regulators publish guidance. None of these turns a case into a rule that binds the next node. They are artefacts without precedential effect. The mechanism that makes precedent work is not the record-keeping. It is the binding, and the binding is a property of the network, not of the individual case.
Precedent is a fallback engine
The registrar who froze was holding two conflicting oaths, with no mechanism to resolve the conflict at the moment of crisis. The network had no fallback for that state.
Law does. When two rules conflict in a legal dispute, the conflict is not left to the judge's interior to resolve from scratch. There is a body of prior cases that have already dealt with the conflict — or with conflicts close enough to be instructive. The judge is not deciding blind. She is deciding in the presence of every previous decision that touched the same seam.
That is a fallback. Not a fallback algorithm — it is not deterministic. A fallback body. A prior resolution the node can reach for when the current case does not have an obvious answer. The fallback does not decide the case. It tells the node that the case has been seen before, in some form, and how the last node resolved it.
The registrar had nothing like this. He had his training, his oath, and the patient. No accumulated body of resolutions for the conflict between the two. No mechanism that had been refined by prior freezes. No way to know that other doctors had stood in the same room, with the same two rules firing, and what happened when they acted — or did not act.
Which is why the same freeze recurs. Medicine has guidance, protocols, and guidelines. It does not have case law. It does not have a mechanism for turning one freeze into a rule that binds the next.
Law holds all three oaths at once
The three-oath problem from the previous piece is not unique to medicine. Law has its own three oaths. Practitioners — solicitors, barristers — are bound to the client. Managers — court administrators, chamber managers — are bound to the operation. Leaders — the judiciary, the law officers — are bound to the system's sustainability and to the public interest.
These oaths conflict, in law, in exactly the ways they conflict in medicine. A barrister's duty to her client can conflict with her duty not to mislead the court. A judge's duty to decide fairly can conflict with the system's need for throughput. A chief justice's duty to the long arc of the law can conflict with the present government's operational demands.
What law does that medicine does not is hold the conflicts in a form that is cumulative, retrievable, and binding. The conflicts are not resolved once and forgotten. They are pinned in the doctrine. This is how the duty to the client yields to the duty to the court, in this class of case, for these reasons. The next time the conflict arrives, the previous pinning is available. Over time, the network accumulates a working map of the seams.
Which is why the manager's oath is not operationally dominant in law the way it is in medicine. It cannot be. The manager's oath in a court is bounded by the practitioner's oath in the doctrine, and the doctrine is visible. It has artefacts — reported cases, commentaries, texts. The manager cannot claim that throughput overrides the duty to the client, because the doctrine says otherwise, and the doctrine can be cited.
Medicine has no equivalent. The doctrine that would bound the manager's oath is not written anywhere, because medicine has no mechanism for writing it. The practitioner's oath is held in the interior of each practitioner, invisible to the manager's layer, and therefore unable to bound it.
Law is not a solution — it is a managed version
It would be easy to turn this into a proposal. Adopt case law for medicine. Build a precedent engine. Let doctors cite prior decisions.
That is not what this piece is claiming, and the reason matters.
Law's mechanism has its own costs. Precedent is slow — slower than legislation, slower than guidance, sometimes slower than the harm it is meant to prevent. It entrenches. Old decisions bind new cases in ways that can outlast their usefulness, and reversing a precedent requires more than a good argument; it requires a case, a court, and the right composition. Law's selection mechanisms are their own — judges are promoted for the same visibility reasons everyone else is, and the elder wisdom of the bench is filtered by the same filters. The doctrine is not free of invisible layers. It has them, in the form of legal reasoning that is too subtle to be captured in a headnote, and the subtlety is often the point.
So the claim is not that law has solved the three-oath problem. Law has managed it — partially, slowly, at cost, and with its own recurring failures.
What law has that medicine does not is a mechanism. Not a solution. A mechanism. A way of turning a case into a resolution, and a resolution into a retrievable artefact, and an artefact into something that binds the next node that faces the same seam.
The registrar's freeze was not a failure of the registrar. It was a failure of a network that had no mechanism for turning a freeze into a rule. Law has that mechanism. Medicine does not. Which means the same failure can happen in medicine for as long as medicine lacks one.
What a precedent engine could do
This is where the dry contribution becomes specific.
A precedent engine would not be case law. It would not be a judicial system. It would be the structural property that medicine currently lacks: a retrievable body of prior resolutions, indexed by the shape of the conflict, available at the point of decision.
The engine would not decide. It would not judge. It would not replace the practitioner's oath. What it would do is hold the accumulated resolutions in a form the practitioner could reach for when the two oaths fired against each other — and could see what happened last time, and the time before, and could know that other nodes had stood where he is standing now.
Which is the same claim that the dry systems made in the earlier pieces, applied to a specific case. The dry system does not supply the wisdom. It supplies the visibility. The practitioner still holds the oath. The network still runs on the manager's oath by default. But the practitioner is no longer alone with the conflict, because the prior resolutions are available, in the same way the doctrine is available to a judge.
The engine would also do something else. It would make the conflicts visible to the network as conflicts. If a particular rule-conflict recurs a thousand times a year, that is not a thousand individual freezes. It is one structural problem, appearing at scale. The engine would be able to say so. Not to resolve it, but to name it — and to name it to the network, where the naming can become a doctrine change rather than a thousand incidents.
Which is the piece's real claim. Not that medicine should adopt case law. That medicine lacks a mechanism for turning recurrence into learning, and that the mechanism can be built. Not to replace the practitioner's judgement. To supply the thing the network has never had: a memory of its own conflicts, in a form the node can reach for at the moment of the freeze.
What comes next
The last piece in Care is about that engine — what a hybrid network would actually do at the point of a freeze, and how the dry system would route around the conflict rather than resolve it.
The Absolution Router is the closing design — not a solution to the three-oath problem, but a specific structure the network could hold so that the freeze stops happening to the same node, alone, every time.