There is a set of phenomena that are usually filed under different shelves: Charles Bonnet syndrome, the “Invisible gorilla” of Simons and Chabris, phantom pain and Melzack’s neuromatrix, Ramachandran’s mirror therapy, dreams and their resemblance to hallucinations, introspective blindness and confabulation.
On the surface, these belong to different disciplines: ophthalmology, cognitive psychology, pain neuroscience, somnology, psychiatry. But if you look at them not as “curiosities of the brain” but as manifestations of the architecture of an adaptive system, the same structural movement becomes visible.
The brain does not behave like a truth optimizer. It behaves like a navigator of regime continuity.
Where the external data stream is poor, contradictory, or fails to pass through attention, the system does not “stop”, does not acknowledge the void, does not switch off to zero. It continues the regime. Sometimes at the cost of error. Sometimes at the cost of a phantom. Sometimes at the cost of narrowing the future. But it continues.
The Void Is Inadmissible
Charles Bonnet syndrome is one of the cleanest examples. A person’s visual input degrades, and instead of “seeing less”, the system begins generating complex visual images. Importantly, this is not necessarily psychosis and not necessarily a loss of insight. It can be an almost “technical” consequence of sensory deprivation: when the input channel thins, internal generation does not disappear — it surfaces.
Sleep shows the same thing, only in its normal form. There is almost no external stimulus, yet a perceptual world is constructed — and constructed convincingly enough to be experienced as reality. This is not a failure of “optimization”. It is a regime in which the system maintains experiential continuity without relying on external correction.
Phantom limb and phantom pain are an even harsher demonstration. There is no input at all — the limb is physically absent — yet the body map continues to operate. And sometimes it is precisely this persistence of internal structure that creates a conflict experienced as pain. Here it is especially visible: the absence of input does not mean the absence of experience. On the contrary, the internal mechanism of regime continuation can become dominant.
This is the first architectural pattern: the system prefers generation to void. It does not say “there is no data — therefore there is nothing”. It says “the regime must continue”.
Truth Is Secondary to Admissibility
The “Invisible gorilla” experiment is often retold as an amusing trick about inattention. But structurally it is much deeper. A stimulus can be objectively present and even obvious, yet fail to become part of experience. The reason is not that “the system poorly optimized truth”. The reason is that there is an admissibility gate: attention as a gate that decides what enters the regime of conscious perception at all.
First comes admissibility, then comes content processing. If admissibility did not fire, there is nothing left to optimize after the fact: the event never became part of the trajectory of experience.
This is deeply uncomfortable for any naive picture of the mind as a “truth scanner”. And entirely natural for a picture of the mind as regime navigation.
Hence the second pattern: truth by itself does not guarantee entry into a regime. Entry is guaranteed by admissibility, and admissibility is a separate structure.
Drift Is Inevitable and Cannot Be Cured by “Zeroing”
If you look at these phenomena over time, something else becomes visible: the system does not just choose to “continue” once. It accumulates the consequences of continuation. Prolonged regimes, sustained mismatches, chronic conflicts between expectation and absent input lead to plastic reorganizations, to stable patterns, to the consolidation of a new internal landscape.
This is drift: irreversible structural displacement under sustained interaction and regime continuation.
The naive idea of “suppressing drift” works poorly precisely because drift is not equivalent to error. Error can be corrected through feedback. Drift is a change in the very structure of continuation. If you try to clean it out directly, what you typically get is either latent accumulation of degradation (symptoms seem to quiet down while restriction of the future grows), or rigidity (the system holds itself in a narrow corridor and ceases to be adaptive), or regime breakdown.
Hence the third pattern: drift is inevitable, and it cannot be reduced to “an error that needs to be zeroed”. It can only be navigated.
Identity Is Preserved Even Under Error
Phantom limb shows that the identity of the body schema can be preserved despite gross physical mismatch. Sleep shows that the identity of the observer can be preserved in the absence of the external world. Hallucinatory phenomena show that continuity of experience can be more important than accuracy of stimulus correspondence.
In these regimes the system can err in content but preserve itself as a regime. If we translate this into the language of architecture, what emerges is an uncomfortable but powerful claim:
identity is not equivalent to truth. Identity is the structural continuity of admissible continuation.
Two Failures and One Stable Regime Between Them
Now the most interesting juncture. If we accept that the system has an admissibility gate, then a question arises: what happens if you remove that gate, and what happens if you make it a fully causal instrument?
If you remove the gate, the system gets uncontrolled hallucination. Internal generation has nothing to “press against”. There is no structural boundary holding the regime in form. Generation sprawls, the regime disintegrates, coherence collapses. This is not necessarily “psychosis” in the clinical sense, but it is always a breakdown of containment: too much is admissible, and therefore nothing holds.
If you make the gate fully causal — that is, turn it into an object of direct exploitation, optimization, and manipulation — the system becomes rigid. It begins to game the boundary. It begins to avoid and suppress everything that interferes with comfortable passage. It stops learning and stops taking risks. In the limit, you get paralysis: perfect controllability of admissibility at the cost of losing liveness.
And here is where that intermediate regime appears — the one I call Navigational Cybernetics of order 2.5: permitted visibility of boundaries without causal exploitation.
The system can see constraints, can take structural conditions into account, can recognize where it is. But it cannot turn those conditions into an optimization lever, cannot turn the boundary into a geometry along which to build detours and strategies.
This is not “more control”. It is a prohibition on one kind of control, because that kind of control destroys the long horizon.
Where This Connects to My Architecture
An important note: neurophenomena do not prove the architecture. They do not “confirm theorems”. And I am not going to play the pseudoscientific trick of “look, the brain does it this way — therefore we are right”. That is the wrong move.
The right move is different. I describe an architectural class, and the neurophenomena are particular realizations in which that class can be read empirically: as a system where continuity of admissible regime comes first, and error minimization comes second.
Neuroscience shows that a real, powerful adaptive system does in fact live this way: it maintains the regime even when the data is poor; it holds identity even when the content is erroneous. It does not expose its gates as a simple map available for optimization; it pays an internal price for sustained mismatch and narrows its future possibilities.
And if this is so, then an uncomfortable but honest formula emerges: the brain is not an error optimizer. The brain is a navigator of admissible regimes under finite internal time.
I am not claiming that the brain implements my architecture. I am claiming that the observable phenomena make it very difficult to sustain any simplification in which the mind reduces to truth optimization or error correction.
And it is precisely this gap that the architectural theory of the long horizon closes: the primacy of admissibility, the inevitability of drift, the preservation of identity as a structural regime, and the necessity of an intermediate level of reflexivity where visibility is permitted but exploitation is forbidden.
In short, this entire set of phenomena shows one and the same pattern:
a system is not obligated to be truthful in order to be continuous, but it is obligated to be continuous in order to remain a system.