Can a court take your children from you because of the party you vote for?
The church you attend?
The political beliefs you hold?
Under ordinary constitutional law, not merely because of the belief or association itself. There must be some legally sufficient connection between it and the welfare of the child.
Can a court perform the same transformation — without openly declaring the reason?
No. Making an impermissible decision rule covert rather than overt does not make it permissible.
Does that change when the label attached to someone is “QAnon”?
Again, no. The label cannot do the evidential work by itself. There must be some proven nexus between the category applied to a person, that person’s actual conduct, and whatever legally cognisable danger the court says it is addressing.
The reason I am analysing the work of Judges Norris and Ray in North Carolina is that I believe this may become one of the most significant test cases in America.
For the sake of argument, assume that “Q” is eventually validated, and that “QAnon” is shown to have functioned as an enemy classifier used to deprive Americans of their rights.
Under that hypothesis, this case ceases to be merely a custody dispute. It becomes evidence of a hidden social and cultural conflict—and potentially an unusually clean trace of how such a classifier operated in a pre-Restoration world.
I am building towards the point where the starting concept of this series—plagium—comes full circle.
A republic, by its nature, rejects the forceful dominion of one person over another. That obscure Roman concept from Part One therefore anchors something much larger than an argument over parenting arrangements.
Part One identified a single load-bearing judicial conversion: a permanent residence became something ambiguous. That conversion helped sustain a narrative under which an accomplished change in possession could become the starting state from which custody was subsequently computed.
Because the implications were serious, I published a companion AI analysis explaining the systems concepts behind that finding.
Part Two expanded the experiment.
When the off-ramps are all blocked
Why should you pay attention to a forensic dissection of a court case you were not involved in?
Instead of examining one load-bearing conversion from the bench, I identified twelve materially different branching points. Together they allow us to examine the court as a state-conversion engine: facts enter; judicial transformations are applied; legal and factual states emerge.
If the outputs follow from the inputs, there may be nothing remarkable to explain.
If materially different inputs undergo different transformations yet repeatedly produce the same directional output—even where some of those inputs point elsewhere—then we have a phenomenon worth investigating.
That observation does not tell us what caused it.
Indeed, that is precisely the point.
Because the stakes are so high, I am now offering a second AI companion: not to decide what happened, but to show how we can reason rigorously from the pattern without assuming its cause.
The method is deliberately conservative. Start with the least controversial explanation. Give it everything it can explain. Examine what remains. Add another explanatory variable only when the residual requires one.
The objective is not to find the darkest explanation.
It is to find the least dark explanation capable of accounting for the whole record.
I hope this companion helps make clear why I am spending so much time on what can appear, at first sight, to be the minutiae of one American custody case.
If the issue is merely bad judging, the method should eventually tell us so.
If it is something else, the same discipline should make that increasingly difficult to hide.
PS — I wrote this introduction, not AI, but crisped it up with tech help. I hope you find what follows useful in understanding what is really going on in family courts. This isn’t ultimately about me or my Substack. It’s about your family life—and what happens when the state acquires the power to classify first and justify later.
Part One gave us a control experiment. The farmhouse was unusually useful because it was a comparatively hard factual object. It could not change its mind, become alienated, or decide that North Carolina felt more like home.
Part Two performs a different experiment.
Instead of following one variable through the judicial machinery, it identifies multiple places where that machinery could materially have changed direction.
The result is twelve substantially different branching points, including the farmhouse from Part One. All twelve preserve or further entrench the inherited trajectory.
That sounds dramatic.
But 12/12 is not actually the important finding.
1. Twelve out of twelve proves surprisingly little
A judge can rule against one party twelve times and be completely right twelve times.
One litigant can simply be wrong, dishonest or dangerous. If the evidence points consistently in one direction, we should expect the decisions to do the same.
So perfect directionality is not itself evidence of misconduct.
The experimental result is subtler:
The inputs vary.
The transformations vary.
The output direction does not.
That is what requires explanation.
2. Part One tested a fact; Part Two tests a function
The farmhouse experiment asked what happened to one relatively fixed proposition as it passed through judicial processing.
Part Two varies the inputs.
Jurisdiction is not residence. Residence is not an AirTag. An AirTag is not schooling. Schooling is not passport control. These variables require substantially different kinds of reasoning.
Yet the article finds them repeatedly transformed in ways that preserve the same inherited state.
The object of investigation has therefore changed.
We are no longer merely testing a fact.
We are testing the transformation function.
3. Not every alternative is an off-ramp
This matters because otherwise twelve branches could simply be twelve alternatives selected after seeing the result.
Part Two imposes three tests.
An off-ramp must have been available: genuinely open to Ray on the record and law before him.
It must have been material: capable of affecting something important.
And it must have been consequential: taking it would have interrupted, weakened or reversed the inherited trajectory.
That makes the branching exercise potentially falsifiable.
An imagined alternative is not an off-ramp.
4. The residual matters more than the ruling
Begin with the strongest ordinary explanation:
Ray applied conventional best-interests reasoning.
Give that model everything it can explain. School matters. Friends matter. Stability matters. A teenager’s preference matters. Disrupting an established life can cause genuine harm.
Then inspect what remains unexplained.
Those unexplained differences are residuals.
The method is therefore:
fit the least controversial model
→ inspect the residual
→ add another explanatory variable
→ inspect again.
The residual, rather than the outrage, tells us where to look next.
5. Real evidence can still be manufactured evidence
Part Two distinguishes exogenous from endogenous facts.
An exogenous fact exists independently of the intervention and can therefore help justify it.
An endogenous fact is generated, at least partly, by the court’s intervention itself.
The child really can acquire friends, enjoy school, adapt to North Carolina and become more attached to her new circumstances.
Nothing needs to be fabricated.
The problem is that the temporary arrangement may have helped manufacture every one of those perfectly genuine observations.
So another forensic rule emerges:
Do not ask only whether evidence is true.
Ask how it came to exist.
6. The intervention can alter both sides of the comparison
The asymmetry goes further.
Keeping the child with the father can manufacture evidence of his stability: school, friends, routine and continuity.
Separating the child from the mother can simultaneously manufacture evidence of her instability: travel, disrupted contact, temporary accommodation and a deteriorating relationship.
One intervention can therefore improve the measured characteristics of one alternative while degrading those of the other.
The resulting comparison may look observationally neutral.
Its inputs are not necessarily independent.
7. This is how a ratchet appears
Now introduce time.
Temporary possession generates circumstances. Circumstances generate adaptation. Adaptation generates stability. Stability increases the apparent cost of restoration.
The increased cost of restoration then becomes another reason not to restore.
Nothing in that loop requires conspiracy or fabricated evidence.
It requires only that consequences generated in one direction be allowed to accumulate as independent reasons for continuing in that direction.
Hence Part Two’s central observation:
Time is not neutral when only one outcome is allowed to accumulate the benefits of time.
8. The AirTag reveals antecedent displacement
The AirTag gives us a different kind of conversion.
A concealed tracking device is placed in the child’s luggage. The mother discovers it. Conflict follows.
Ray himself identifies the critical fact:
“You wanted to know where she was.”
The father then acknowledges that he wanted the ability to locate the child.
Yet the eventual Order principally records the mother’s reaction to discovering the device.
The antecedent conduct has not disappeared factually.
It has disappeared analytically.
That yields another portable test:
When a reaction becomes evidence,
inspect what happened immediately before it.
9. Recognition without propagation is more interesting than ignorance
The AirTag also gives us something stronger.
Ray did not merely fail to understand its purpose.
He identified it himself.
Part One supplied the same phenomenon with the farmhouse. Contrary information entered the system, was made explicit, and yet the eventual representation did not update accordingly.
That suggests a useful diagnostic category:
recognition without propagation.
A system that never receives corrective information may simply be ignorant.
A system that recognises corrective information yet repeatedly preserves the state that information might disturb presents a different problem.
10. “Where we are now” contains hidden history
“Where we are now” sounds like a neutral starting point.
Sometimes it is not.
If the present state was produced by the very process under examination, then present reality is partly an output being fed back into the system as a new input.
Its history has not vanished.
It has been compressed into the present state.
Start there and restoration can look like disruption.
Start before the disputed transition and exactly the same movement can look like restoration.
The baseline changes the sign.
11. Formal availability is not functional availability
This may be the most important lesson of the off-ramp metaphor.
A legal system can formally permit reconsideration. A temporary order can formally be interlocutory. Contrary evidence can formally be admissible. A judge can formally possess discretion to change course.
None of that establishes that correction is functionally available.
If materially different corrective inputs repeatedly enter the system without changing its trajectory, the distinction between formal and functional availability becomes observable.
A road can contain twelve exits on the map while still functioning as a one-way system.
12. None of this requires conspiracy
Status-quo bias may matter. Judges may hesitate to disturb another judge’s arrangement. Child-welfare doctrine naturally values stability. Delay generates new facts. Institutions resist rollback.
Those mechanisms could produce strong directionality without anybody secretly selecting an outcome.
That is why twelve blocked off-ramps do not tell us why they were blocked.
They tell us that the question has become worth asking.
And they give us a disciplined way to ask it.
13. The off-ramps are test vectors
The twelve branches are therefore more interesting as experiments than as grievances.
Change the variable. Change the evidence. Change the legal issue. Change the causal structure.
Then observe what the system does.
This resembles system identification: perturb a system in substantially different ways and infer characteristics of the hidden transformation from its responses.
Part One supplied a calibration object.
Part Two supplies multiple test vectors.
Together they begin to reveal the machinery.
14. Adjudication has a topology
Part One was principally about states.
Part Two is about paths between states.
A corrigible system permits:
contrary information → reconsideration → rollback.
A ratcheted system tends towards:
contrary information → reinterpretation → commit.
Both systems may contain exactly the same formal laws, motions, hearings and judicial discretion.
What differs is the topology of correction: which information is actually capable of changing which existing states.
That may be Part Two’s deepest systems lesson.
Formal availability of an off-ramp does not establish functional availability of an off-ramp.
What Part Three now has to discover
Part One found an anomalous transformation in a hard factual variable.
Part Two asks whether directional behaviour replicates across substantially different variables.
The answer is disturbing: twelve materially different branching points, twelve outcomes preserving or further entrenching the inherited trajectory.
But neither result establishes motive.
That is the next problem.
Start with ordinary best-interests adjudication. Measure the residual.
Add anchoring and status-quo bias. Measure again.
Add institutional resistance to rollback. Measure again.
Only then introduce candidate hidden variables and ask whether they explain observations that the simpler models leave behind.
The objective is not to discover the darkest explanation.
It is to discover the least dark explanation capable of accounting for the whole record.
If ordinary judicial error does that, stop.
If it does not, the residual tells us where to look next.
And that is where Part Three begins.