Thinking

This book builds from the bottom up. First the mechanism — how a single neuron operates, how networks produce consistency, what the filter does, how constitution shapes processing. Then the application — how autism, ADHD, and other psychological conditions arise from this mechanism.

The following overview contains the core of each chapter in compressed form. Reading it gives you the model in outline. The detail chapters fill out the arguments.

Chapter 1: Neurological Foundations

A theory of thinking starting from the neuron. The central thesis: neural networks are consistency machines. They minimize inconsistency between signals, and this single mechanism explains what thinking is, how models of the world arise, why truth is reachable, and why more input brings both load and depth.

Neurons Like Steadiness

A neuron is a cell that excites other cells when its own excitation level crosses a threshold. It lives in a stream of signals from other neurons and produces signals itself.

The fundamental property: neurons like steady signals. Predictable input runs stably and cheaply. Uneven input — jumps, breaks, contradictions — costs energy and forces the network to adjust its connection strengths until the uneven signal is predicted as steady the next time around. This is Hebbian learning: what fires together wires together, patterns become expectation. Smoothing means turning surprise into routine.

Networks Are Consistency Machines

A single neuron smooths its input. A network of many neurons smooths between many sources at once.

With one signal, consistency is trivial: whatever comes in is the truth. With two signals, a common explanation must be found in which both signals run smoothly. With many signals, a network of cross-constraints emerges in which every signal limits all others. Finding a configuration that smooths all signals together is hard — but when it stands, it is not arbitrary. It has been checked from many sides.

Learning is this smoothing process over time. Thinking is the same process in real time. What we call "model of the world" is the current connection configuration in which the network processes its inputs consistently.

The network reaches its smoothness actively, through prediction: every level predicts what the level below should report. If the prediction is right, silence reigns. If it deviates, correction kicks in — either the model is adjusted, or the deviating signal is discarded as noise. What becomes conscious is the unexpected, not the arriving.

Approaching Truth

Consistency alone is not truth. A model can be internally consistent and still wrong — it only has to smooth the signals the system happens to have.

Hence the second thesis: More signals from more domains push the model closer to truth. A model that explains only visual input can be arbitrary. The same model also has to explain auditory input, then tactile, then social, then temporal consistencies. With each domain, models drop out that break against at least one. What remains is a thin path compatible with all of them. This path is truth.

Truth is the convergence point of many independent consistency requirements. Cross-domain synthesis is truth-approximation. A system that processes more signals can reach deeper truths than one that processes fewer — provided it can carry the consistency load.

More Signals, More Load, More Depth

The central asymmetry. More signals mean more cross-constraints, and more cross-constraints mean a tighter model. But consistency work is costly. Every inconsistency is tension in the network, and every tension costs energy and reorganization.

A system with few signals lives in smooth prediction: comfortable, but thin. A system with many signals lives in continual tension: uncomfortable, but potentially deep.

Which configuration a network runs depends on two factors: how many signals come through (sensor sensitivity) and how much consistency work the network can perform (synthesis capacity). The thin-skinned with high synthesis becomes a pattern-weaver — many domains integrated into one model, exhausting but truth-near. The thin-skinned with low synthesis hits overload — inconsistencies cannot be resolved, the system is torn apart, and stimulus avoidance becomes necessary. From outside the two look alike. Inside, one weaves; the other drowns.

Three Paths to Consistency

A network that senses inconsistency has three strategies:

1. Update the model. Connection strengths are redistributed until the new signal fits. This is learning.
2. Reshape the world. Instead of adjusting the model, the environment is changed to fit it. This strategy comes in all sizes — from setting up a small routine, to tidying the room, to building large works. Whoever writes, programs, builds, designs, or arranges their daily structure is doing it.
3. Strengthen the filter. Instead of changing model or world, the internal processing strength is raised — the filter is consciously turned on. Meditation, breath work, exercise, sleep, and vagal regulation belong here: they raise top-down precision and reduce tension in the network without changing input or model.

Most systems mix all three. Which mixture dominates depends on constitution, capacity, and environment. Updating the model expands the inside. Reshaping the world expands the outside. Strengthening the filter raises the carrying capacity for both.

Definition

A compact definition follows. Thinking is the activity of a network producing consistency between signals. It is the basic operation that carries both perception and action. Perceiving is consistency with incoming signals. Acting is consistency with intended signals. Insight is the moment a new cross-constraint smooths many things at once.

A network that thinks well can take in many signals from many domains and integrate them into a consistent model. Whoever only receives, without smoothing, drowns. Whoever only smooths, without receiving, dreams.

Chapter 2: The Filter

So far the network has been described as if all signals enter on equal footing and are processed by it. That is not how it works. Every functioning brain has a mechanism that decides which signals are passed upward at all — and dampens most of them already at the bottom. This mechanism is the filter.

How the Filter Works

The filter sits inside the hierarchy itself. Higher processing levels continually send predictions downward to the level below. These predictions do two things: they say what the next signal should be, and they dampen all signals that match the prediction. What the higher level expects is turned down at the lower level.

Practically: the hum of the fridge is predicted, so it barely reaches consciousness. The clothing on the skin is predicted, so the pressure signal is smoothed away. The peripheral visual field is roughly predicted, so a sketch suffices.

Important: information is not destroyed by this. It is compressed. The lower levels still process the full raw information; only what confirms the prediction does not rise further. What the higher level receives is the difference between world and expectation plus the abstract concepts computed below. The filter subtracts and adds simultaneously.

The Filter Is an Abstraction Hierarchy

Dampening is only half the filter's work. The other half is abstraction.

Every level of the hierarchy produces predictions for the level below. The prediction "there should be a lamp at the upper left" dampens pixel signals matching the lamp and delivers the intermediate result "lamp, upper left" as a finished concept to the level above. That is how seeing a lamp arises rather than processing a pixel matrix. Low levels deliver lines and edges, middle levels objects, high levels meanings, intentions, concepts.

The filter does the same thing on the social level. From facial micro-movements, vocal fluctuations, and posture it computes "friendly" or "tense" — a concept that arises in parallel with sensory detail processing. Whoever has a functioning social filter sees pixels and person and mood at once. The lower levels are not switched off; they keep delivering, but the result is abstracted to "friendly person" rather than arriving above as a data stream.

A functioning filter is a hierarchical translator, not a mute switch. It does not make the world poorer but richer: the same input is translated simultaneously into several layers — pixel, object, meaning, intention, mood. With higher sensor precision (as in HSP), all these layers are computed more densely, not only the lower ones. HSPs feel more emotion in a room, not less — the filter renders the social signal at higher gain.

What is missing in autism is reliable abstraction together with a full raw stream. The lower material is there and pushes upward, but the abstract concepts are formed less decisively. The person is seen, but "friendly" must be computed from individual signals rather than arriving as a finished concept.

Why the Filter Exists

The filter is the answer to a quantity problem. A brain receives millions of sensory data points per second. Passing them all upward at full precision would be energetically impossible. Instead, sorting happens at every level: the expected stays down, only the unexpected rises.

That is efficient, but it has a price. The filter shows the model of the world minus its corrections, not the world itself. Whoever has a good filter sees a predictable world — even when the world might be more surprising than the model admits. The filter optimizes for energy, not for truth.

The Filter's Adjustment Parameters

The filter is not binary. It has two axes, and on each axis systems can be calibrated differently.

Sensor precision (bottom-up): how strongly raw signals are weighted before they meet the prediction. High sensor precision means small deviations already arrive — the world pushes through. Low sensor precision means only coarse stimuli get through.

Top-down precision (top-down): how strongly the prediction dampens the input. High top-down precision means the model smooths decisively — the world is seen as it is expected. Low top-down precision means the filter barely engages, everything comes through raw.

The two axes are independent. Their combination decides what arrives at the system:

• High + high: sharp perception with strong expectation smoothing. Efficient, but blind to model deviations.
• High + low: thin skin, no filter. Everything comes through, everything has to be processed. Sensory-overload profile.
• Low + high: dim perception, strong model. The model overwrites what little arrives. Hallucination profile.
• Low + low: little input, little smoothing. Energetically cheap, cognitively flat, dampening-near.

Most people sit in the middle of both axes. But the edges are where it gets interesting — that is where the constitutions and conditions sit that make up the diversity of the population.

What the Filter Does for Consistency

Updating the model and reshaping the world act on conscious prediction errors. The filter acts before that — it decides which signals become prediction errors at all. Strengthening it raises carrying capacity without changing model or world. Weakening it provides more raw material for learning at the price of constant prediction tension.

The filter is therefore the invisible main factor deciding how much work a brain expends per day. Two people with the same synthesis capacity but different filter strength live in two different worlds. One finds life simple. The other finds it full.

Chapter 3: Constitutions

So far the network has been described abstractly — sensor sensitivity, top-down precision, synthesis capacity as variables that can vary. They do not vary at random. They sit in the body, in the autonomic nervous system, and they are individually calibrated. To understand a person, you have to know their bodily constitution.

Three States

The autonomic nervous system can occupy three basic modes, balancing one another. In the terminology used here:

The three states are not morally ordered. Activation carries action. Clarity carries learning and connecting. Dampening carries recovery. A healthy system moves flexibly between them, with clarity as default. Dysregulation arises when one state chronically dominates.

Each state directly moves the parameters. Activation cranks sensor precision for threat and reward signals upward and narrows the attentional window. Clarity balances sensor and top-down precision so the network stays learning-ready. Dampening lowers sensor precision overall — the system shuts down, less comes in, less is processed.

Three Constitutions

Which state is reached how quickly and how deeply depends on inborn constitution — how sensitively each component is calibrated, how easily a system jumps between modes, where it tends to go by default.

Three axes, each tied to one ANS component:

Pure types are rare. Most people are mixtures, with a dominant and a secondary axis. The mixture shapes character long before experience comes in.

States Move the Parameters

The direct effects on the parameters come from states, not from constitutions.

Constitution Shapes Access

Constitutions affect the parameters indirectly: they decide which states a system tips into easily and how intensely it operates there.

Constitution Is Not Destiny

Constitution sets the default but not the permanent state. An agile system under chronic stress lives in activation, even if its default tendency is clarity. Real life plays out in the interaction between constitution, environment, and the three paths.

Whoever knows their constitution knows which strategies actually work for them. An inert system gets duller from stimulus reduction, not clearer — it needs measured stimulation. A reactive system gets overwhelmed by more stimulation, not more productive — it needs filter strengthening. An agile system gets stabilized by routine, not caged — it needs structure to use its jumping mode at all.

In Chapter 4 onward, the psychological conditions are mapped onto this. They are points in the constitution space that suffer in the average environment — because their default settings do not fit the world they were born into. They are not diseases in any qualitative sense.

Chapter 4: Autism

Autism, in the model built up here, is a specific configuration of the parameters, not a disease. Constitutionally reactivity-dominant (often with an agility component), with a chronically weak filter and high sensor precision. The system is built to take in many signals — and it does.

The Configuration

Three parameters characterize the autistic phenotype:

1. Sensor precision high. Raw signals from all domains arrive at full intensity. What neurotypicals get filtered out as background — the hum of the fridge, the tag in the shirt, the flicker of the light, peripheral movement — comes through unchecked in autists.

2. Top-down precision low. The filter barely smooths. Expected signals are not reliably turned down. The automatic abstraction hierarchy is missing: raw material is less often translated into finished concepts, more detail stays visible.

3. Update reward focused on consistency. Model updates are sought, but within chosen domains. The drive is depth within the existing, not jumping to the new (that is ADHD). Inconsistency in the own model is unbearable; consistency is actively produced.

This is the HIPPEA hypothesis (High Inflexible Precision of Prediction Errors in Autism, Van de Cruys et al. 2014) in this book's language: prediction errors are reported too precisely, the filter does not dampen them away, the system has to work through every single one.

The Experience

Subjectively: the world feels complete, not quiet. The hum of the fridge is audible. The tag in the shirt rubs continuously. The light flickers because the 50-Hz modulation is actually seen. Other people wonder what the fuss is about — they do not hear the hum because their filter has turned it off.

From this follows almost everything described as autism symptomatology:

• Consistency drive — every model update costs, so predictability is actively sought. Routines are world-reshaping: the environment is built so that it produces few prediction errors.
• Social exhaustion — human-human contact carries the highest predictive uncertainty, because every other person carries their own hidden model.

High-Functioning and Overload

Three quantities decide whether an autistic system becomes productive or torn apart: synthesis capacity, input volume, and setting fit.

With high synthesis capacity in a matching setting, the rich input becomes productive. The system integrates the many signals into an increasingly dense model. Cross-domain constraints accumulate, the model gets deep, truth-approximation runs. This person becomes a pattern-weaver: high-precision perception, deep synthesis, special domains penetrated.

The classical image — the programmer, the physicist, the mathematician — is a consequence of this configuration, not a cliché. Whoever as an autist works every day in the same domain with clear rules gives their filter exactly what it needs: a stable setting where synthesis capacity can unfold without being overwhelmed by ever-shifting raw signal. Sheldon Cooper is the caricature of the phenotype. In self-structured research or programming environments, such people become some of the most productive minds of a generation.

Setting fit is its own quantity. The same system fails in an open-plan office with constant phone ringing, micro-politics, and shifting demands. The environment delivers too much inconsistent signal to be forced into a common model — the synthesis capacity is there but cannot get used.

There is a too-much of signal. Even at the highest synthesis capacity, input load is bounded. Whoever lives in an environment that produces more inconsistency than the network can smooth per day does not reach high function — no matter how high the capacity. Small-scale world-reshaping (stimulus reduction) is then a precondition for synthesis to even get going — not a weakness.

With low synthesis capacity, the same rich input becomes overload — independent of setting. Inconsistencies cannot be resolved, the network is torn apart. The only strategy is drastic small-scale world-reshaping: tight action radius, strong routines, clear rules. This protects but constrains world-access massively.

So three phenotypes: the pattern-weaver in a matching setting, the overwhelmed high-capacity person in the wrong setting, the low-capacity person in unavoidable reduction. From outside the latter two look alike but are different problems. A change of setting helps one and not the other.

Filter as Intuition

What runs automatically through the filter in neurotypicals — facial expression becomes "friendly," small talk becomes "normal," pixels become "lamp," the whole stream gets translated into finished concepts — is nothing other than intuition. Intuition is the speed at which a model delivers answers without explicit thinking. It is the filter in action.

Autists have less intuition. What others feel, they have to derive. What others experience as "obvious," they have to compute from individual signals. That is slower. It is more exhausting. And it is, at first, a disadvantage.

But: what has been worked through logically is understood more deeply than what only functions intuitively. Intuitive understanding is fast but does not check itself — the filter says "fits" and the matter is settled. Logical understanding is slow but forces every component to justify itself. Whoever has to walk every step themselves builds a model along the way that is markedly more robust than the automatic concept.

So autism is at first a disadvantage — less intuition, slower in everyday life, social friction. In the right setting the same disadvantage becomes an advantage: what others only feel, the autist can make explicit. What others take for granted, he can examine. Complexity that others can no longer intuitively penetrate becomes accessible to him because he has to process complexity explicitly anyway. The path is harder at first, deeper later on.

Social Complexity

Human-human interaction is the hardest domain because it imposes several consistency loads at once:

• Other people have hidden states that have to be modeled to predict behavior
• One's own presence changes their states recursively
• In groups, multiple models are simultaneously active, plus their interaction network
• Social signals (facial micro-movements, vocal modulation, posture) are processed as raw data points at low filter strength, instead of being smoothed into "friendly"

What neurotypicals condense automatically through the ventral-vagal filter into a simple social reading has to be evaluated manually by autists. It is possible but costs more energy per encounter than reading a book. Family is a particular load, because known models have to be held in parallel with complex drama dynamics.

Conclusion

Reactivity as constitutional core explains the sharpness of high-functioning autism: special interest, deep dive, low tolerance for sloppiness. With agility added, the AuDHD phenotype emerges — sharpness plus jumping attention, the pattern-weaver in pure form.

Autism is the configuration that takes in more signals than the average filter accommodates — and that therefore has to do without automatic intuition. In a fitting environment with developed synthesis capacity, the forced detour through explicit thinking becomes the most productive configuration for truth-approximation. The difficulty lies in the mismatch between a high-precision system and a world optimized for medium filters — not in autism itself.

Chapter 5: ADHD

ADHD is a different axis from autism. Where autism is shaped by high sensor precision and a weak filter, ADHD is shaped by a strongly elevated update reward. The system is heavily rewarded for processing prediction errors — new patterns, highly informative signals, switches are attractive, almost compulsive.

Constitutionally ADHD is agility-dominant: fast state switches, high mobility between modes, sensitive ventral vagus. The system jumps easily from one mode to the next — and that is precisely the mechanism that makes ADHD productive and exhausting at the same time.

The Configuration

Three parameters:

1. Update reward very high. Every prediction-error gradient is strongly rewarded. The system orients automatically to wherever the most can be learned. Highly informative signals magnetically pull attention.

2. Sensor precision moderate to high. Not as sharp as in autism, but broad. The attentional window switches quickly, covers many domains over the course of a day.

3. Attentional focus unstable. The filter itself is not primarily disturbed — it works, but it is constantly being re-aimed at new signals. As soon as somewhere a higher information gradient appears, the reward system pulls attention there. The filter then turns the previous away and the new sharp — until something more informative appears again.

What looks from outside like a concentration weakness is, inside, a highly sensitive update system constantly jumping to the strongest prediction-error source in the environment. Whoever has ADHD in a world full of highly informative stimuli cannot do anything but jump.

Hyperfocus and Drift

The apparent inconsistency — concentration weakness and hyperfocus at the same time — resolves once you understand update reward as the main mechanism. Hyperfocus appears when an activity is highly informative and delivers new prediction errors second by second: a game with dense feedback loops, a programming problem, a creative track that constantly opens new connections. The system stays because it is constantly rewarded.

The moment the activity becomes boring — little new information per unit of time — the system jumps away. Routine tasks, administration, repetition are unbearable, because the reward system finds nothing to hold onto — not because the system couldn't do them.

Co-occurrence with Autism

Autism and ADHD often co-occur. The model makes clear why: they run on independent axes.

• Autism: high sensor precision, weak filter, focus on consistency within chosen domains
• ADHD: high update reward, jumping to novelty

The two axes can be combined freely. AuDHD (autism + ADHD) yields a phenotype with precisely perceived raw signals from many domains and high reward for processing them. When the world is sensorily rich and learning is strongly rewarded, the system jumps constantly between pattern zones — exhausting but productive for cross-domain synthesis.

That is the pattern-weaver in pure form: many signals from many domains, high synthesis capacity (through the reactivity component), and a system that jumps between domains (through the agility component). Truth-approximation across breadth. Precondition: a setting that allows the jumping without sanctioning it as weakness.

Pure ADHD without Autism

Pure ADHD without an autism component looks different. Perception is not necessarily high-precision, the filter is not generally weak. The system is agile-jumpy, creative, broadly interested, but not thin-skinned in the autistic sense. Classical: many hobbies, quick to enthuse, quick to bore, hard to motivate for boring duties, surprisingly productive when the topic carries.

Strength is mobility and innovation. Weakness is stability — what was planned as a major project one day is uninteresting the next, because the system has already jumped on.

Setting and Strategy

What ADHD needs is stimulation with structure, not less stimulation. An empty open-plan office with routine tasks is the worst environment — low information content, no natural hyperfocus possible, the system drifts constantly. A rich environment with clearly structured challenges is optimal: programming, research, creative projects, any job with a dense feedback loop.

Reshaping the world means for ADHD: finding activities in which the reward system collaborates. Filter strengthening (exercise, breath work, sleep) reduces jump frequency. Model updating happens constantly anyway — that is the strength, not a deficit.

ADHD is a configuration that initially produces friction in a world built for steady attention. School, office routine, slow duties presuppose that attention is held on the uninformative — exactly what this reward system is not built for. In a fitting setting, the same high update reward becomes a drive for innovation and rapid cross-domain connection — properties that are above-average valuable in a complex world.

Chapter 6: Further Conditions

If thinking is producing consistency, most psychological conditions are shifts of the parameters far enough from the middle that the system suffers in the average environment. They sit on the same parameter space as all other configurations, not in a separate "sick" category.

What follows are sketches — the basic mechanic of each condition in this book's language, not full descriptions.

PDA — Pathological Demand Avoidance

PDA is a defense module sitting on top of an existing constitution — not an axis of its own. External demands are read as attempts to lay foreign predictions over one's own model — model corruption from outside. Reflex reaction: refusal, to protect model integrity.

The reflex itself is universal. How it manifests depends on which cognitive level the system is currently operating on.

ODD is the childlike form: direct defiance, "no because no," emotional opposition to group pressure. This is the fast, intuitive reaction every child knows. Neurotypical adults fall back to this level easily under autonomy pressure — they have the path from childhood, and stress reactivates it.

PDA is the refined form: resistance with a justification apparatus, internally present even when external compliance would be possible. The system can mask, but only at high cost.

Autists hardly fall back to the childlike defiance level even under stress. The mechanism follows from the missing intuition: the autistic processing system has no fast intuitive-emotional fallback path. Defiance is an intuitive group reaction, and that very path is not default for autists. The system either keeps going explicit — even under load, because simple thinking feels wrong when the intuition that carries it is missing — or tips into dampening. It does not throw a tantrum like a child, because tantrum requires the intuition it lacks.

PDA is therefore not simply "the autistic form of ODD." It is the same autonomy-defense reflex, executed on a higher cognitive level — more differentiated, harder to counter with anti-tantrum strategies, but also less easily broken by group pressure.

OCD — Obsessive-Compulsive Disorder

OCD is a prediction-error loop. A specific concern — the stove might be on, the hands might be contaminated, a particular thought might be bad — produces a prediction-error signal that cannot be silenced. The person checks, sees "all clear," but the internal update does not get through. The error signal stays active, so a check is repeated. And again.

In the model: the sensor evidence is not fed into the model at all, because the top-down prediction "danger" is so precise that it overwrites the reassuring signal. Instead of updating the model, the system repeats the check, hoping at some point to gather enough evidence. That is a model update that does not close — the second of the three consistency strategies, locked in an endless loop.

Schizophrenia

Schizophrenia is often described as the opposite of autism — top-down too high, sensor too low. That is too simple. Both conditions share problems with precision weighting; they only tip at different points.

The shared mechanism is gap-filling. Everyone does it. Under uncertainty the hierarchy automatically produces predictions to close the gaps, because a gap means unresolved inconsistency. A vague shape becomes a person, a sound becomes a voice, a piece of behavior becomes an intention. That is normal.

In high autism, gap-filling is used actively: high sensor precision plus weak automatic filter create gaps everywhere, and the consistency drive demands that they be filled. The system produces hypotheses — what the person might be thinking, what is behind the behavior, how system X might function. That is pattern-weaver mechanics. Crucially: the hypotheses remain testable. External signals correct the filled model when they come in.

Schizophrenia is the pathology of the same mechanism. The filling runs through without external signals correcting it. Hallucinations are the positive symptoms — the brain produces content through prediction alone, and the bottom-up stream is not weighted strongly enough to refute what was generated. Delusion is the same dynamic on a semantic level: an internal model that no longer lets reality signals correct it.

Somatically, the same state shows up as global neural hyper-synchronization: the integration system forces too many levels into the same prediction, instead of staying open to correction. What looks in the filter model like overly strong top-down precision is, on the system level, hyper-arousal of the highest integration layer.

The transition zone is real. Highly autistic systems with little outside correction — isolation, missing resonance partners, one-sided information sources — can tip toward schizotypal self-runaway, not because one is the other, but because the same mechanic loses its connection to the correcting world under different conditions.

Depression

Depression is giving up. The system sees no path that brings the perceived signals into a consistent solution and shuts down. The filter is turned up so as few signals as possible come through — the only available consistency is "flat." Update reward drops, sensor precision for positive signals drops, energy expenditure is reduced. Dampening state, dorsal vagus dominant.

Constitutionally, classical depression appears often with inertia dominance or as the end state of chronic reactivity exhaustion (lived too long in activation, tips into the opposite).

Agitated depression is a second mode: activation and dampening simultaneously. When the noise gets too large for the filter to hold, the sympathetic nervous system runs up in parallel without the dampening releasing. Gas and brake at once. This is the system's worst state, because none of the usual paths work: dampening would need activation, which is already there and helps nothing; activation would need calming, which collapses into dampening. Resources are spent on two opposing processes simultaneously. Classic burnout pattern, very hard to leave.

The consistency drive does not give up; the system gives up. What appears as drivelessness is the exhaustion of the update system after working too long against unresolvable inconsistency.

Anxiety Disorders

Anxiety disorders are chronic activation with a threat model. Sensor precision for threat signals is turned up, the model predicts many states as threatening, the system stays in sympathetic mode. Constitutionally reactivity-dominant, often with disrupted routines that prevent re-settling into clarity.

The prediction "threat" is constantly confirmed by the elevated sensor precision — anything not safe is dangerous. The system stays in mode because its own vigilance behavior keeps feeding the threat model. A self-sustaining loop on the constitutional level.

PTSD

PTSD is a prediction error that could not be integrated. A single event was so large that the model could not absorb it into its structure — the raw trace stays unprocessed, isolated, stored at full precision. With trigger signals it is reactivated at full intensity, as if the situation were now.

Normal smoothing through repeated confrontation does not occur, because the update signal is too large to be let through — the system closes before every attempt to integrate it. Therapy is essentially controlled re-encounter in a state where the update actually goes through.

Borderline

Borderline is an unstable model of other people. Social micro-signals are read with high precision, but the models built from them are not stable. A person is "the dearest" and hours later "the betrayer" because a single new signal flips the entire model. High sensor precision for social signals plus low stability of top-down models — the filter does not integrate over time.

The mechanic partly overlaps with the autistic phenotype: high social sensor precision plus weak filter that should stabilize the models of others. The difference lies in phenotype. In classical BPD the social model flips between idealization and devaluation, emotionally charged, often with abandonment panic and unstable self-identity. In the autistic variant the swinging is more cognitive: "I don't know how to place this person, they are behaving differently today than yesterday," without the love-hate jump — and the consistency demand on the self (integrity drive) is the opposite of BPD-typical identity diffusion.

It is worth distinguishing the two mechanisms. Whoever has high social precision and a weak filter will recognize parts of the BPD description without having clinical BPD. The shared denominator is the axis; the different endpoint is the pathology.

Psychopathy

Psychopathy is low sensor precision for the emotional signals of others plus high top-down precision for one's own models. What neurotypicals experience as automatic empathy — mirroring the model of other persons — is missing. The own model stays egocentric, without the sensors correcting it. Other people appear as objects with useful functions rather than as their own consistency carriers.

What This List Shows

There is no fundamental jump between "neurotypical" and "disordered." All conditions are points in the same parameter space. What makes a configuration into a disorder is the mismatch with the environment and the extent to which the system suffers under it — not the position itself.

Healing is therefore movement in the parameter space — strengthening the filter, correcting the models, reshaping the world, letting constitution work with constitution rather than against it. "Back to the norm" is not the goal. Suffering is real, therapy is meaningful, but the picture "broken vs. healthy" misses the mechanism.