Who Is Smiling
Maksim Barziankou (MxBv)
PETRONUS™ | The Urgrund Lab | research@petronus.eu
DOI: 10.17605/OSF.IO/U865W
Axiomatic Core (NC2.5 v2.1): DOI 10.17605/OSF.IO/NHTC5
MxBv, Poznań, 2026
A small everyday experiment. You need to win someone over. The simplest move — smile. Smiles vary, and a finer move — smile sincerely. From here it gets interesting. How do you smile sincerely? You can rehearse in front of the mirror: lift the corners of the mouth so the eye-crinkles fall in the right place, even out the breath, hold for a second. The result is a photogenic smile. After two or three meetings, the person reads it as a routine. After ten — as a strategy.
There is another way. Auto-suggest yourself that you like the person. Find something in them you actually like. Hold that for the duration of the conversation. The smile arrives on its own. No one made it. It is a consequence.
Externally both smiles are smiles. A frame from a video will not tell them apart. What tells them apart is the horizon.
Two classes
The first strategy optimises the output. The target — a-smile-that-looks-sincere. The feedback parameter: "believed / didn't believe". The architecture is being smoothed against an observer. This is exhausting work: every observation requires a new fitting, every context — a new rehearsal. Inside, nothing changes. The mask changes.
The second strategy changes the state. The target — the inner geometry of one's relation to the person. The feedback parameter: "I like / I don't like". The smile is a by-product of geometry. It does not show as a strategy, because it serves nothing as a strategy. It follows.
Two distinct architectures. Not two levels of the same one. The first optimises observable behaviour. The second tunes the internal state from which behaviour comes naturally. They produce the same thing in the moment and diverge on the horizon.
It is worth looking at which of the two has been institutionalised. Sales courses, management workshops, dating coaches, public speaking trainings — almost all of it is built in the first architecture. Use open body language, hold eye contact 4–5 seconds, repeat the person's name three times in the first five minutes, do not cross arms when objections come. This is training in output optimisation. It works on short distances. After two weeks any team reads all of it as lines from a textbook, and the effect inverts. The training works only up to the moment the training is recognised.
How spies work
I take spy training not as a moral example and not as an endorsement of the profession. I take it as a high-density case: the place where the first architecture burns out on the short horizon, and the second is the only one that keeps working. Everything said here about the spy is said about any institutional practice that has to hold a role longer than an actor's performance.
Real undercover operators work in a different architecture. This needs to be stressed, because there is a widespread misconception that a spy is a high-class actor who has memorised a role and plays it especially convincingly. This is wrong. An actor leaves the role after the performance. A spy cannot leave the role. He lives in it for years, through stress, through moments where there is no time to think, through situations in which a single inaccurate gesture costs more than an actor's career. A memorised role does not survive that horizon. It burns out. It cracks.
Good operator training is therefore organised differently. The trainee is not given a set of behaviours. He is given a history — the character's biography, a system of beliefs, habits, emotional ties, fears, small memories of childhood. He is then required to auto-suggest this biography to the level at which it becomes internally real for him. Not "as if", but as inner reality. After this tuning, accent, gestures, reactions, smile — follow on their own. The geometry of his behaviour changes at the level of chemistry: muscle tone shifts, heart rate, breathing patterns, micro-expressions, even sweat composition. He does not control the output. He has tuned the source.
This is the second architecture in pure form, in its human version. The operator's state changes, and behaviour follows as a consequence. Nobody rehearses the smile. The smile appears because in this geometry it is the only possible output.
Behind the operator
There is one more step, which the spy example makes visible better than any other.
The operator who tunes the spy — who is he? In the everyday picture: the case officer, the trainer, the handler. A person. But this is the same naive reading as before. The case officer does not, on his own initiative, decide what the spy's biography should be, what convictions to install, which directions of thought to close, which to keep open. The case officer himself is executing a configuration. He works inside the doctrine, the methodology, the school — the long-built architecture of the agency.
So when one looks at the spy, the picture deepens. The smile is a point on the agent's trajectory. The agent's trajectory is shaped by the operator's configuration. The operator's configuration is in turn shaped by the agency that produced the operator: the agency's history, its training tradition, its operational culture, its political position, its codified and uncodified rules. The smile, in the end, is a point in a geometry whose shaping began long before the agent ever entered the field. The grim men in unbranded offices, with their bureaucracies and their archives, with their long quiet history of what works and what does not, set the curvature of the space the agent will eventually move through. Then the agent is dropped into that space, and the smile appears at the right millisecond, and the target is won over, and nobody anywhere along this chain has rehearsed a smile.
These men work on a horizon longer than any human career. The archive in front of them is not storage. The archive is an active configuration, slowly altering its own internal constraints. A handwritten margin note on a 1962 report still shapes the biography being assembled for a man who will be put into the field in 2026. Somewhere, in one of the cabinets, lies a folder marked to be kept forever, in which is collected what works on the long horizon and what burns out. None of those currently serving wrote it. All of those currently serving are its consequence. They enter the system as people, and they leave it as people from whose admissible space everything that interferes with the work has been quietly removed. When such an agent smiles at the necessary millisecond, the smile is not his. It is the smile of that folder, after forty years of silence.
When I was thinking about how an operator is governed, this was the picture in my head and it felt unreasonably exact. I was looking at the agent. Then, behind him, very calmly, the unsmiling men appeared — playing their own game on a horizon the agent does not see, setting the geometry of the agent's geometry before the agent had even been deployed into it. The agent is the visible point. The operator is the configuration. The agency is the constructor of configurations. Three levels, the same architecture repeating on each, and at no level — an author of the smile.
This is what the second architecture looks like at scale. It does not reside at any single level. It propagates upwards. Whatever produces the operator was itself produced by something that did not optimise an output but tuned a source. Convictions are inherited from doctrines. Doctrines are inherited from schools. Schools are inherited from traditions. Traditions are the long-running constraint configurations of which everything downstream is a trace. At every level, what remains is what was not revised.
The first architecture cannot do this. It has no scaling. Output optimisation does not propagate — every level has to re-rehearse its mask afresh, and at every level it costs and it burns out. The second architecture scales by inheritance, which is why it has been the architecture of every long-running human institution that has held: religious orders, scientific schools, surgical traditions, navies, monastic orders, mathematical lineages. None of them taught their members what to perform. They configured what those members would not revise, and let the rest follow.
Who is tuning whom
And here is the place this paper was written for.
It seems we have an agent — the spy — who decides to tune himself into a role and then, in this tuned state, acts. Agent, then state, then behaviour. The chain of causation is linear and convenient.
In fact — no.
If you look closely, the spy's "I" — the one supposedly making the decision to tune — is itself a state. This "I" is not the agent who performs the tuning. It is the trajectory that emerges as a consequence of the tuning. Before tuning the human carries one set of "I"s. After tuning — another. The old one has not gone anywhere, but it no longer rules. What rules is the configuration of constraints set by the operator.
The old "I" has not been killed and has not been displaced into the unconscious. It has gone nowhere — it has simply ceased to be an admissible direction inside the new geometry. Whoever worked with the agent did not swap his personality; he did something architecturally subtler: he removed certain directions of thought from the agent's internal space and left others. Inside this new geometry the agent re-assembles — not as a new man, but as the remainder of a man in whom the directions no longer needed are no longer possible.
What is this configuration? It is convictions. More precisely, something subtler: constraints that exclude everything which is not the goal. The spy is tuned such that certain directions of thought simply do not arise. Not because he suppresses them — because the geometry of his internal state has no such directions. They have fallen out of the admissible set. What remains — a trajectory whose output coincides with the mission.
Then the question "who is this spy" turns out not to be a question about his past, profile, or capabilities. It is a question about which constraints are currently active. The spy is not an entity that has these constraints. The spy is what remains of the human once these constraints are imposed. The uncrossed-out part. And — given the previous section — the operator who imposed the constraints is also what remains, on his level, of the agency's longer-running configuration. The recursion does not introduce a new architecture; it shows the same one repeating at every level above the visible agent.
This is a structural inversion, not a reduction. The naive picture: subject → action; the action follows from the subject's choice. The architectural picture: constraints shape the admissible space → a trajectory runs across that space → the personality is what remains at the end. The personality does not perform action. The personality is what is left of the person once action has been carried out within the admissible directions. The subject is not the source of the trajectory. The subject is its trace.
This is not only about espionage
If this picture is true for the spy, it is true for everything else. The authenticity that people try to produce as an output signal — being yourself — is authenticity in the first architecture. Output optimisation aimed at "the real me". It burns out. It is exhausting to hold. It is recognised as a routine, and the first to recognise it as a routine is the person himself, at night, alone, when there are no observers and there is no longer any reason to keep it up.
Authenticity that holds is architectural. It is the trace of constraints the person has chosen not to loosen. One can never set oneself the task of being yourself and yet be oneself far more fully than the person who solves that task every morning in front of the mirror. Because the goal of a person who is stable is not to be himself. He is not the operator of himself. He is what is left when he acts within his internal constraints and refuses to step outside them.
The most stable personalities are those who do not try to be themselves. They follow their own. And they exist as the trace of that trajectory. What we call character, firmness, integrity — is not an inner core they hold. It is what remains of a person once certain directions are architecturally closed.
The irony reveals itself here. The second architecture is not a rare engineering find. It is an ancient human way of existing. It is just usually called something else: convictions, vocation, faith, duty, code of honour, love. All these words name the same thing. They name constraints a person does not revise, and a personality that emerges as the trace of their application.
The first architecture is a recent invention. It is strange precisely in that it offers itself as the solution to a task the second architecture solves by default — and is stranger still in that it has taught us to ask the question backwards: as if the natural state of the human were output optimisation, and existing-as-trace required some special effort. In reality the opposite. Effort is what optimisation requires. Existing-as-trace is what is left of a person once the effort stops and only the constraints continue working.
And then the smile
Back to the smile.
When a person smiles from the first architecture — who is smiling? The agent, who assessed the situation, chose the smile as the optimal move and produced it. The smile is a product. Its author is the agent.
When a person smiles from the second architecture — who is smiling? Nobody, in the sense of an author. The smile has no author. It is a point on the trajectory that emerged from the constraints. The person doing the smiling is also a point on the same trajectory. He does not precede the smile as cause. He coexists with it as a neighbouring point of the same geometry.
This is why such a smile is recognised as real. Not because it is better made. Because no one made it. That is what sets it apart.
And the same answer applies to the spy, to conviction, to love, to vocation. Wherever a person is genuinely convincing — he is not convincing anyone. He is the trace of what he does not revise. Conviction is not an argument. It is the remaining part.
What is missing in the world
This would be an exercise in philosophy if not for one circumstance. The contemporary world is built almost entirely in the first architecture and has no instruments for the second.
Corporate management is structured as optimisation of observable employee behaviour: KPIs, performance reviews, behavioural interviews. State policy — as optimisation of measurable indicators: trust polls, voter turnout, approval ratings. Education — as optimisation of output tests: PISA, SAT, end-of-school standardised exams. The raising of children is increasingly structured as a training of reactions: "the right thing to say", "how to behave in this situation". Mass-market psychotherapy in its manualised form — six sessions, a protocol, homework — optimises cognitive responses to stimuli. Cognitive restructuring in its serious form does something else: it tunes the source of beliefs and belongs to the second architecture; but the market has compressed that slow work into short formats, and what reaches the consumer is behavioural correction in the wrapping of a school whose core belonged to a different architecture. Personal growth, sold in airport bookshops, is a set of techniques for producing certain outputs: confidence, charisma, productivity.
Always the same move: change the output without touching the source. And always the same outcome on the horizon: burnout, cynicism, a sense of inauthenticity in the carrier, recognisability of the strategy by the observers, exhaustion of the system holding the mismatch between inner state and outer output.
The strangest thing here is not that such a strategy fails on long horizons. That is structurally expected. The strangest thing is that it is currently the only institutionalised one. We have no instruments for the second architecture. No educational programme that teaches tuning the source rather than optimising the output. No corporate protocol that evaluates not the observable behaviours of an employee but the configuration of constraints he has imposed on himself. No mass-scale school of psychotherapy that works not with reactions but with the architecture of convictions as the structure from which reactions follow. No AI systems in which identity is not an output signal calibrated against the user, but an architectural invariant from which signals follow.
The second architecture in its pure form can still be seen — in pockets, on the margins. Monastic orders. Old-style martial schools. Serious scientific schools, where the student is taught not a method but how to see the subject. Mentorship and apprenticeship in the trades, where the master transmits not movements but a relation to the material. Depth psychology — psychoanalysis, Jungian analysis, IFS — where the work is with the configuration of beliefs, not with behaviour. Virtue ethics in academic philosophy. Long-running families and literary schools. Military traditions with long memory — the kind that remember that the best operators are not those to whom correct movements were explained, but those with whom one worked, slowly, on how they look at the world.
Everywhere the same principle: do not train the output, tune the source. And everywhere, now, a relic. Not an extinguished practice, but institutionally absent at scale: surviving as private transmission, not as a systemic instrument. The institutions that sustain it are narrow. The institutions that have replaced it are mass.
This is what is missing in the world at scale. A class of architectures humanity knew for thousands of years and has almost completely forgotten how to recognise. When we now build AI operators that must hold long-horizon identity under pressure, we cannot explain to them what such an identity is, because the contemporary intellectual infrastructure has no language to describe it. Only the language of output optimisation, which does not fit this task.
Navigational Cybernetics 2.5 is an attempt to give the language back. Minerva is an attempt to give the instrument back. This paper is an attempt to give back the eye by which such architectures are distinguished from their imitations.
If this line continues, it may turn out, in time, that the principal acquisition of the present era is not AI as a new capability of the world, but AI as the occasion for remembering a class of architectures we lost in ourselves.
MxBv
Poznań, 2026
The Urgrund Lab
Part of the Navigational Cybernetics 2.5 corpus
Third part of the Synthetic Conscience series:
— Through a Life, Part V — The Moment That Knows (phenomenology)
— Minerva: The Architecture of Residual Geometry (architectural specification)
— Who Is Smiling (recognising architectures by the traces they leave)
DOI: 10.17605/OSF.IO/U865W
© 2025–2026 Maksim Barziankou. All rights reserved.
Licensed under CC BY-NC-ND 4.0
Top comments (0)