Forem: oleg kholin

The Evolution of the 3D Printing Problem: From Technological Optimism to Structural Deadlock

oleg kholin — Sat, 11 Apr 2026 11:08:43 +0000

The development of 3D printing over the past decades has been accompanied by a persistent expectation of its inevitable mass adoption. The logic appeared straightforward: the technology matured, hardware became cheaper, materials became widely available, and software gradually improved in usability. Within this framework, it was assumed that further cost reduction and simplification would eventually make the 3D printer as common a household device as a paper printer or a microwave oven.

However, the actual trajectory has been different. Despite technological maturity and accessibility, 3D printing has not become part of everyday domestic life. This discrepancy between expectation and reality is often explained through familiar arguments: the lack of a “killer use case,” high barriers to entry, poor economic competitiveness compared to mass-produced goods, and inferior product quality. Yet these explanations remain superficial and fail to address deeper structural causes.

The Initial Misframing: False Universality

The core issue begins with how the question itself is framed. The assumption that any mature technology must become mass-market ignores a fundamental distinction between classes of tasks. Some technologies serve daily or regularly recurring needs, while others address rare, highly variable, and context-specific problems.

Low cost and accessibility are not sufficient conditions for mass adoption. There are many examples of inexpensive, highly capable devices that never become household standards because they do not correspond to everyday needs. The ability to use a tool does not imply the necessity of using it.

In this context, 3D printing was incorrectly positioned from the outset. It was treated as a potential mass household technology, whereas by its nature it belongs to the category of specialized tools—similar to equipment used in workshops or production environments.

Reframing the Context: From “Every Home” to “Every Workshop”

Correcting the framing leads to a different interpretation. A 3D printer is not a household appliance in the conventional sense. It is a tool designed for solving problems that arise irregularly but require a high degree of customization.

From this perspective, it becomes clear that the technology has already found stable domains of application. Jewelry production, custom components for technical devices, advertising and promotional items, and educational construction kits all demonstrate effective use of 3D printing. These domains share a common characteristic: small batch sizes, high variability, and the absence of economic justification for traditional industrial manufacturing.

Thus, the issue is not the absence of demand, but its nature. The demand is not mass-market—it is niche, yet stable and reproducible.

The Illusion of Technological Limitations

Many arguments against broader adoption of 3D printing rely on outdated assumptions. Claims about insufficient precision, strength, or functionality increasingly fail to reflect current reality. Modern desktop systems are capable of producing working mechanical components suitable for practical use without additional finishing.

Other limitations, such as water resistance or consistency of output, are often interpreted as inherent to the technology. In practice, however, these depend heavily on process parameters. Their resolution lies in standardization and reproducibility of settings, not in altering the underlying physics of the process.

Thus, many perceived “limitations” are not technological but infrastructural.

The Ecosystem as a Consequence of Task Structure

Another commonly cited issue is the lack of a developed ecosystem—unified model libraries, standardized print profiles, and user-friendly tools. However, a deeper analysis shows that an ecosystem cannot emerge independently of a structured understanding of tasks.

In mature engineering and software systems, the primary layer is not the toolset but the ontology of objects and operations. Users work not with abstract geometry, but with entities that have parameters and behavior. This allows systems to scale through extensions, reuse, and accumulation of knowledge.

In 3D printing, the situation is reversed: tools exist, but there is no shared understanding of what tasks are being solved or how. As a result, each user constructs an individual workflow, and accumulated experience does not scale across the system.

Under these conditions, an ecosystem cannot be built directly. It can only emerge as a byproduct of task systematization.

The Representation Problem: From Geometry to Parameters

The dominant model exchange format—static geometric files—limits reuse and adaptability. These models contain no information about purpose, constraints, or functional parameters.

A parametric approach, by contrast, defines objects through relationships and constraints. This enables adaptation to specific conditions without breaking functionality. However, adoption of this approach is constrained by the lack of accessible tools aligned with real-world workflows.

The gap between existing CAD systems and practical user behavior remains one of the central barriers.

The Role of Adjacent Technologies

The evolution of 3D printing is closely tied to the maturity of adjacent technologies. One of the most critical missing components is affordable, accurate 3D scanning. The ability to quickly capture the geometry of existing objects would significantly simplify many practical workflows, particularly those involving replication or repair.

The absence of such tools increases labor costs and reduces accessibility, further limiting adoption. In this sense, 3D printing remains partially constrained by the immaturity of its technological ecosystem.

The Limits of Generative Solutions

Attempts to compensate for the lack of models through generative approaches encounter a fundamental limitation. Generative systems are oriented toward creating new forms, while many real-world tasks require accurate reproduction of existing objects under functional constraints.

Without embedded engineering logic, generated models may appear plausible but fail in practical use. This highlights the distinction between form synthesis and engineering design. The former may assist the latter, but cannot replace it.

The Absence of a Dominant Use Scenario

Another defining feature of 3D printing is the absence of a dominant, unifying application scenario. In successful technological domains, development is typically organized around a small number of clearly defined use cases, which drive standardization and infrastructure.

In contrast, 3D printing is characterized by a wide range of fragmented applications without consolidation. This fragmentation hinders standardization and slows ecosystem development.

The Non-Obvious Cause: The Absence of a Risk-Bearing Actor

The deepest layer of the problem lies in the distribution of risk. Building a fully functional ecosystem requires long-term investment, coordination across multiple layers, and acceptance of uncertainty. Yet the benefits of such an ecosystem are distributed across many participants, while the costs are concentrated on whoever initiates it.

Hardware manufacturers are incentivized to protect proprietary advantages rather than standardize. Software companies focus on high-margin enterprise markets. Open-source communities lack the resources to deliver robust, production-grade systems. Investors are reluctant to engage with long-term, uncertain, and weakly monetizable opportunities.

As a result, no actor emerges for whom building the ecosystem is a rational decision. This creates a structural deadlock: the technology exists, demand exists in niches, partial solutions exist—but integration does not occur.

This distinguishes 3D printing from cases of successful technological scaling. In those cases, there is always an actor for whom the cost of inaction exceeds the cost of building the system. That actor may be a company, a consortium, or a public institution—but it exists.

In 3D printing, such an actor has not yet emerged. Moreover, the current distribution of incentives actively discourages their appearance. Benefits are diffuse, while risks are concentrated.

Therefore, the absence of an ecosystem is not the root cause, but a consequence. The root cause lies in the economics of risk. As long as the cost of integration exceeds its expected return for any individual participant, systemic solutions will remain unrealized.

The Resulting Picture

3D printing is not a failed mass-market technology. It is a mature tool for a specific class of problems that do not align with everyday consumer use.

Its limitations are not primarily technological, but structural:

incorrect framing of mass adoption as a goal;
absence of a formalized task space;
inadequate model representation formats;
mismatch between tools and real workflows;
immaturity of adjacent technologies;
lack of dominant application scenarios;
absence of an actor willing to bear integration risk.
Future Directions

The future of 3D printing depends less on improving hardware and more on advancing the organization of knowledge and systems around it:

developing a clear taxonomy of tasks and use cases;
transitioning from geometric to parametric models;
creating tools aligned with actual workflows;
standardizing print profiles by object type rather than hardware;
advancing accessible methods for geometry acquisition;
identifying a limited number of scalable application domains.

Until such developments occur, 3D printing will remain an effective but localized tool—widely used in professional and semi-professional contexts, yet lacking a mechanism for broader systemic adoption.

Rising on the Shoulders of Giants: Top 10 Technologies That Succeeded by "Hijacking" Infrastructure

oleg kholin — Fri, 10 Apr 2026 04:58:47 +0000

Introduction: Why Do Some Technologies "Fly" While Others Stagnate?
According to the "Four Trees" framework, the most expensive and time-consuming stage of progress is growing the "Roots"—Trees 3 & 4 (the auxiliary tools and the infrastructure for mass production). Most brilliant ideas perish because they attempt to grow these roots from scratch.
However, there exists a strategy of "Engineering Refactoring." Instead of growing its own tree, a Strategic Subject performs a scan of the global market and identifies a mature Tree 4 in a completely different industry. By "hijacking" this existing infrastructure and adapting it to a new task, they achieve a lightning-fast leap to Tree 2 (the mass product). This approach allows them to collapse the cascade of costs by 80–90% and seize the market.
Below are 10 examples of how identifying and utilizing "foreign" auxiliary trees created new technological empires.

Consumer Drones Tree 1 (The Idea): An autonomous flying robot. The Donor (Tree 4): The smartphone industry. The Refactoring: DJI and others performed a "reconnaissance from the small," assembling drones from "spare parts" intended for phones. This collapsed the cost of such devices from $100,000 to $1,000.
Artificial Intelligence (AI/Neural Networks) Tree 1 (The Idea): Machine learning via neural networks. The Donor (Tree 4): The gaming industry (GPUs). The Refactoring: AI stagnated for decades until developers "recognized" in NVIDIA’s gaming video cards the perfect tool for matrix calculations. AI "parasitized" the global demand for video games, gaining a ready-made computational base.
Electric Vehicles (Tesla) Tree 1 (The Idea): A mass-market, high-performance electric car. The Donor (Tree 4): The consumer electronics industry (laptops). The Refactoring: Elon Musk used standard 18650 lithium-ion cells already mass-produced for laptops. Tesla rose on the "roots" planted by Panasonic and Sony, which were already mature and cheap.
Uber and Modern Logistics Tree 1 (The Idea): Real-time global management of movement. The Donor (Tree 4): Military satellite navigation (GPS) + 4G networks. The Refactoring: Uber utilized the existing Tree 4 created by the Pentagon and telecommunication giants. By changing the mechanism—using an app on a "borrowed" Tree 4—they annihilated the old taxi industry.
Medical Express Tests (PCR/COVID Tests) Tree 1 (The Idea): Instant molecular diagnostics at a scale of billions. The Donor (Tree 4): The food and beverage industry (PET bottle production). The Refactoring: The industry "hijacked" the production of PET preforms (the blanks used for soda bottles). The same mass-production lines of Tree 4 allowed the world to be flooded with cheap tests.
"Digital Twins" in Architecture Tree 1 (The Idea): Photo-realistic modeling of cities and factories. The Donor (Tree 4): Game Engines (Unreal Engine / Unity). The Refactoring: Hijacking game-engine technology allowed architects to design factories with a speed and fidelity that traditional CAD systems could never match.
Warehouse and Domestic Robots (LiDARs and Sensors) Tree 1 (The Idea): Robots that navigate space autonomously. The Donor (Tree 4): The automotive industry (ADAS systems). The Refactoring: Robotics "harvested the fruit" from trees planted by Toyota and Mercedes, which collapsed the price of LiDARs and cameras by implementing them in mass-market cars.
Cloud Computing (AWS) Tree 1 (The Idea): Selling computational power as a utility. The Donor (Tree 4): Amazon’s retail infrastructure. The Refactoring: Today, half the internet runs on the "roots" Amazon originally grew to support its own online bookstore.
Vertical Farming (AgroTech) Tree 1 (The Idea): Year-round food production within urban centers. The Donor (Tree 4): The LED display industry (TVs and Smartphones). The Refactoring: Vertical farming became viable only when mass production of screens collapsed the price of LEDs. Agriculture "plugged into" the auxiliary tree of the electronics industry.
3D Bioprinting Tree 1 (The Idea): Printing human organs and tissues. The Donor (Tree 4): Office inkjet printing. The Refactoring: The first bioprinters were created by hacking standard HP and Epson printers. The mature Tree 4 mechanics of droplet delivery were hijacked to print living cells.

Investment Radar: Identifying Future Unicorns through Tree Analysis
For an investor, the "Four Trees" model serves as a precision instrument to identify hidden market leaders at an early stage. A true "unicorn" is often born not in a lab, but at the intersection of a mature, external infrastructure and a novel business objective.

"Donor Signal" Detection The beginning of an explosive growth phase for a new player can often be predicted by analyzing the financial reporting of entirely different companies—component suppliers. The Core Concept: A sudden, unexplained spike in sales of specific components among "shoulder companies" (producers of chips, sensors, batteries) often signals the emergence of a hidden giant in an adjacent industry. Example: An abnormal growth in demand for miniature gyroscopes and Li-Po batteries between 2008 and 2010 was the signal for the birth of the consumer drone industry, well before brands like DJI became household names. Investors should watch those who suddenly begin buying "smartphone parts" for non-smartphone purposes.
Investment Arbitrage on the "Broken Cascade" The most profitable companies are those that reach the mass-market product stage (Tree 2) without bearing the colossal costs of building their own fundamental base (Trees 3–4). The Strategy: Look for companies with abnormally high margins in their early stages and a low level of Capital Expenditure (CapEx) compared to industry leaders. The Indicator: If a company shows results comparable to industry giants while having a 10x smaller R&D budget, it is a sure sign they have successfully performed an Engineering Refactoring and are using someone else's infrastructure as a free resource.
Searching for "Universal Roots" (Predicting the Next Wave) An investor can predict the next wave of technological breakthroughs simply by evaluating the maturity of specific "donor markets." The Forecast: As soon as a technology in one industry (e.g., satellite internet, laser scanners, or solid-state batteries) reaches the Tree 4 stage—meaning it has become cheap, reliable, and mass-produced—it automatically opens a "window of opportunity" for new Tree 1 ideas in adjacent fields. The Action: Analyze which mature "roots" are currently sitting on the market, unused. The entity that is first to "graft" these roots onto a new market niche will become the next global leader.

Conclusion: The Lesson of the Strategic Subject
All these examples share a single pattern: Success came not through the invention of new "roots," but through the ability to recognize them in other industries.
The winner is not the one who tries to build everything from scratch, but the one who performs Engineering Refactoring and "grafts" their idea onto the strongest and most massive Tree 4 available on the market. Do not just follow science journals; watch the supply chains of mass-market components. Real business subjectivity is the ability to recognize someone else's mature Tree 4 faster than the competition and rebuild your architecture to exploit it, collapsing the cost structures of everyone else in the process!

The Invisible Roots of Progress: Top 10 Supermaterials Stuck in the Laboratory

oleg kholin — Wed, 08 Apr 2026 20:20:26 +0000

The popular essay "The Four Trees" offers an original lens through which to view technological progress. According to this concept, the development of any technology rests upon four metaphorical "trees":

Tree 1 (The Idea): The fundamental concept or laboratory proof-of-concept. The principle is proven; the physics works.
Tree 2 (The Mass Product): The stage of mass production and widespread infrastructure. What we produce at scale and use in daily life.
Trees 3 & 4 (The Auxiliary Roots): Auxiliary tools and the secondary technologies used to produce them. These are the "hidden" roots—the lithography machines, the specialized furnaces, the methods of purification, and the precise manipulation of matter.

The Great Barrier: From Assembly to Integration

The reason many supermaterials fail to go mainstream is deeper than mere "cost." We are currently stuck in the trap of Miniaturization. This is the stage where we simply shrink individual components and attempt to connect them (similar to how vacuum tubes were replaced by discrete transistors).

The true revolutionary leap is Micro-miniaturization (Integration). This is the transition from "assembling discrete parts" to "forming a structure." In microelectronics, we don't solder millions of transistors together; we grow them simultaneously as a single, integrated structure on a silicon wafer through deposition and etching.

The tragedy of modern supermaterials: We still treat them as "discrete parts" (we try to "cut" graphene or "glue" a nanotube). We are still thinking in the category of "assembly," whereas we desperately need a "lithography for materials." Until we learn to form the structure of a device directly out of the material itself, we will remain in the era of "expensive transistors," never reaching the era of "cheap integrated circuits."

Below are the Top 10 Tree 1 materials waiting for their "integrated revolution."

1. Graphene: Two-Dimensional Carbon

Tree 1 Status: Proven in 2004. A single-atom-thick layer of carbon. The strongest and most conductive material in the universe.
The Trees 3–4 Bottleneck: We are still trying to "transfer" it like a delicate film. This is the era of assembly. For graphene to reach Tree 2, it must be grown directly into the specific regions of a chip as part of a unified integrated circuit, rather than being "pasted" onto existing ones.

2. Nitinol and Shape Memory Alloys

Tree 1 Status: Alloys (e.g., Titanium-Nickel) that return to a complex original shape when heated.
The Trees 3–4 Bottleneck: We currently produce them as discrete "parts" (stents, wires). We lack the technology to integrate "shape memory" directly into the 3D structure of a product during the fabrication stage, allowing the material itself to act as the mechanism.

3. Carbon Nanotubes (CNTs)

Tree 1 Status: Cylindrical carbon structures, 100 times stronger than steel and lighter than aluminum.
The Trees 3–4 Bottleneck: We can produce "nanopowder" (a discrete additive), but we cannot yet form a continuous macro-structure (like a thread or a sheet) without losing their unique properties at the molecular boundaries. We need a method of "weaving" the structure at the moment of formation, rather than attempting to assemble billions of individual fibers.

4. Metallic Glasses (Amorphous Metals)

Tree 1 Status: Metals with a disordered, liquid-like atomic structure. Extremely strong and immune to corrosion.
The Trees 3–4 Bottleneck: We are limited by the requirement of "extreme cooling rates," which restricts us to making only thin ribbons or small parts. We lack the Tree 4 technology to form bulk structures where the amorphous state is preserved during the casting of large masses.

5. Aerogels: "Frozen Smoke"

Tree 1 Status: A crystalline lattice consisting of 99% air. The lightest solid and the world’s best thermal insulator.
The Trees 3–4 Bottleneck: Production requires supercritical drying in high-pressure autoclaves—a boutique, "batch-assembly" method. To become Tree 2, aerogels must evolve into a material that can be deposited like a spray-on foam directly at a construction site, forming its structure in-situ.

6. MXenes: 2D Metals

Tree 1 Status: Two-dimensional crystals made of metals and carbon, capable of ultra-fast battery charging.
The Trees 3–4 Bottleneck: Currently obtained through "etching" (a subtractive and dirty chemical process). This is a discrete, wasteful method. We need the technology to "grow" MXenes directly as electrodes within the pre-integrated structure of a battery.

7. Borophene: Single-Atom Boron

Tree 1 Status: A 2D layer of boron, even stronger and more flexible than graphene.
The Trees 3–4 Bottleneck: It can only survive in an ultra-high vacuum. We lack the technology for "integrated encapsulation"—where the material is grown and immediately sealed with a protective atomic layer in a single, continuous process.

8. Perovskites: "Printable" Solar Power

Tree 1 Status: Crystals that convert light to electricity more efficiently than silicon.
The Trees 3–4 Bottleneck: They degrade rapidly in the presence of moisture. The solution is not just "better chemistry" but a breakthrough in "integrated sandwich-structure" fabrication, where the active perovskite and its transparent protection are formed as a unified, airtight structure during the printing process.

9. High-Entropy Alloys (HEAs)

Tree 1 Status: Alloys made of 5 or more metals in equal proportions, offering extreme heat and radiation resistance.
The Trees 3–4 Bottleneck: The problem of homogeneity. We need "atomic mixing" tools (such as high-speed laser deposition) so that the alloy is formed directly as the final part, rather than being smelted into a bulk ingot that requires further, less precise processing.

10. Synthetic Spider Silk: Bio-Kevlar

Tree 1 Status: Stronger than Kevlar and more elastic than Nylon.
The Trees 3–4 Bottleneck: We can produce the protein (the raw material), but we cannot yet "form the thread" (the structure) with the same molecular grace as a spider. This is the transition from "brewing a soup" in a bioreactor to "molecular-level weaving."

Final Conclusion

Today’s supermaterials are stuck at the transistor level of the 1950s. We have learned how to create them, but we have not yet learned how to unite them into "Integrated Circuits of Matter."

The problem is not that these technologies are inherently too expensive; the problem is that we are still trying to assemble the future by hand, piece by piece, instead of forming its structure as a unified whole. The entity that first creates a "lithography for materials"—moving from the assembly of parts to the integrated growth of systems—will become the new technological leader, controlling real progress!

Decontextualization of Anthropomorphism in Robotics: An Epistemological and Physical Analysis of a Paradigm Shift

oleg kholin — Sat, 28 Mar 2026 19:38:24 +0000

Abstract
Contemporary robotics relies heavily on anthropomorphic morphology as a universal standard for interaction with the human environment. This paper proposes a reconsideration of this assumption. It is shown that anthropomorphism is often not an engineering necessity, but a historical and cognitive legacy that emerged within a specific infrastructural context. A method of decontextualizing anthropomorphic assumptions is proposed through physical stress-tests of the environment, demonstrating the limitations of humanoid architectures. The concept of separating social and physical compatibility of agents is introduced, and a program of adaptive tests is proposed aimed at identifying optimal morphologies for extreme and unstable environments. The paper formulates a framework for transitioning from anthropocentric design to physically conditioned morphological optimization.

Introduction The majority of contemporary robotics is developed within an implicit assumption: a robot must be compatible with infrastructure created for humans. From this assumption, a second is often derived: a robot must be anthropomorphic. However, this assumption is rarely subjected to systematic analysis at the level of the physics of interaction with the real environment. In many scenarios — rescue operations, unstable surfaces, chemically active zones, destroyed infrastructures — humanoid morphology proves to be not optimal, and sometimes physically unstable. This leads to the key research question: can anthropomorphism be not a universal engineering solution, but a contextually limited historical design strategy? This paper proposes viewing the current situation as an epistemological shift in robotics, in which a change in the physical context of tasks naturally leads to the loss of applicability of old paradigms — without the need for their direct refutation.
The Physical Impossibility of Old Paradigms Scientific paradigms sometimes cease to function not because they are logically refuted, but because the physical context in which they were effective disappears. In engineering systems this manifests particularly clearly: a change in environment automatically changes the optimal morphology of the agent. When a robot must function under conditions of:

unstable surfaces,
degrading objects,
chemically active environments,
phase transitions of materials,

anthropomorphic architecture begins to encounter fundamental limitations of mechanics and stability.
In this sense, the paradigm shift occurs not through argumentation, but through the physics of tasks.

The Historical Contextuality of Infrastructure Human infrastructure is often perceived as a universal environment to which all artificial agents must adapt. However, infrastructures were formed historically and adapted to human biomechanics. The history of technology shows that: environment and agents evolve jointly. Examples:

automobiles led to the appearance of asphalt roads and traffic signals,
wheelchairs — to ramps and changes in architectural standards,
electric vehicles — to the infrastructure of charging stations.

Thus, the requirement for robots to fully adapt to existing infrastructure ignores the historical precedent of technological co-evolution.

Biomechanics as a Variable Human biomechanics is often regarded as a fixed model of interaction with the environment. However, real conditions demonstrate the opposite. When the environment changes, the movement strategy changes as well:

on slippery surfaces, additional points of support are used,
on steep inclines, the center of mass is shifted,
in unstable environments, the contact area is increased.

Consequently, the requirement to universally preserve anthropomorphic form means fixing one point in the space of possible morphologies, ignoring the adaptability characteristic of engineering systems.

Social Acceptance Does Not Require Anthropomorphism One of the common arguments in favor of humanoid robots is the convenience of interaction. However, empirical reality demonstrates the opposite. Many technological agents have been successfully integrated into the social environment without anthropomorphic form:

ATMs,
order terminals,
self-checkout machines,
information kiosks.

People interact with them without discomfort, because the key factor turns out to be not resemblance to a human, but:

predictability of behavior
and clarity of functional role.

This allows two independent parameters to be identified:

social compatibility
and physical compatibility.

Physical Limitations of Anthropomorphic Morphology Humanoid architecture possesses a number of characteristics that in certain environments become limitations:

high center of mass
limited contact area
discrete step kinematics
high sensitivity to loss of traction
complexity of load redistribution

A thought experiment involving a humanoid robot moving across thin ice with a load demonstrates these limitations.
Narrow contact points create high pressure on the surface, increasing the probability of structural failure of the support. A high center of mass reduces stability during sliding, and the discrete structure of the step limits adaptation to a continuously changing surface.
Under such conditions, anthropomorphism becomes not merely a suboptimal solution, but a potential engineering error.

The Method of Decontextualization of Anthropomorphism This paper proposes a method that can be described as the decontextualization of engineering myths. The idea of the method is as follows: an engineering hypothesis is tested not only in a standard environment, but also under conditions of changing physical context. If the model loses its operability when the environment changes, its universality proves to be illusory. The method includes three stages:

changing the physical context of the task
observing the degradation of the original morphology
searching for an alternative architecture

In this way, the paradigm is dismantled naturally — through incompatibility with new conditions.

The Evolution of Adaptivity Tests To identify the limitations of anthropomorphic systems, a sequence of tests of increasing complexity is proposed. Test 1 — Retention of a Degrading Object The system must accompany an object that is gradually losing its shape and structure. This tests the ability to adapt to continuous changes in geometry. Test 2 — Evacuation of a Human During Morphological Collapse A human in an extreme environment may lose the structural stability of the body. The robot must complete the rescue before the point of irreversible damage. This introduces a temporal and dynamic component to the task. Test 3 — Extraction of Materials Before Phase Transition The task includes chemical and physical instability of the object, which may become dangerous. Here the ability of the system to operate under conditions of changing matter is tested. These tests shift the focus from the form of the robot to the physics of interaction.
The Precedents Approach Instead of directly asserting a new paradigm theoretically, development through a chain of engineering precedents is possible. First, tasks appear in which old architectures systematically fail. Then different research groups independently find new solutions. As a result, a new practice is formed that gradually becomes the standard. This path is slow, but sustainable: it creates change through the accumulation of facts, not through declarations.
Separation of Social and Physical Architecture One of the key conclusions of the paper is the necessity of separating two functions of the robot:

the social interface
and the physical executive body.

Anthropomorphism may be useful as an interface:

for communication,
for recognition of intentions,
for reducing the cognitive load on the human.

However, physical morphology must be determined exclusively by the physics of the environment.
Separating these levels opens up new architectures of robotic systems.

Co-evolution of Environment and Agents The history of technology shows that environments change under the influence of new agents. In the long term, the emergence of new types of robots may lead to changes in infrastructure:

adaptive surfaces,
dynamic transport environments,
hybrid architectural systems.

Thus, the future of robotics may lie not in copying human form, but in the joint design of environment and agents.

Conclusion The paper proposes an analytical framework in which anthropomorphism is viewed not as a universal standard of robotics, but as a historically conditioned strategy, effective only in certain contexts. It is shown that a change in the physical environment of tasks naturally leads to the loss of applicability of old morphologies. Under these conditions, a transition becomes necessary — from anthropocentric design to physically conditioned morphological optimization. Such a transition is not the victory of one idea over another, but rather a change in the plane on which questions are posed in robotics. On the new plane, the key criterion is no longer the robot's resemblance to a human, but the correspondence of its form to the physics of the environment.

AI and Creativity: Making Sense of a Real Shift

oleg kholin — Wed, 25 Mar 2026 10:25:02 +0000

Our perspective on Tim Green's article "No Consent, No Credit, No Pay"

What Is Actually Happening

The public debate around generative AI and artists' rights is focused on legal details — datasets, lawsuits, licensing models. But behind all this noise, the main point escapes notice: we are witnessing not a technological improvement of existing tools, but the elimination of the very need for intermediaries between an idea and its realisation.

The chain used to look like this: idea → specialist → tool → result. Now it looks like this: idea → result. AI did not replace the designer with a better Photoshop. It made the designer an unnecessary link in the chain. And this is permanent.

Filters and plugins never cancelled the tools themselves. 3ds Max, After Effects, Photoshop — they remained necessary, which meant the people who knew how to use them remained necessary too. AI became a thin client that replaced both the tools and the specialists in one move: designers, layout artists, retouchers, illustrators, pattern makers. You no longer need them — and you no longer need what they used either.

The Analogy Everyone Misses

The authors of such articles and the participants in court proceedings compare what is happening to piracy or copyright infringement on the internet. That is an imprecise analogy.

The precise one is Gutenberg's printing press. It did not give scribes a better writing tool. It made the scribe an unnecessary link in the distribution of text. The profession did not disappear overnight — new niches emerged, new forms of craft. But the economic foundation was undermined irreversibly.

Even closer is the history of photography. The arrival of smartphones with quality cameras did not destroy photographers in direct competition. It simply turned out that the vast majority of consumers were satisfied with the level of photography available on their phones. Professionals survived in niches: artistic reportage, film photography, studio work. Enthusiasts of film still exist — just as enthusiasts of valve amplifiers do. But the monopoly on quality imagery collapsed forever.

AI is doing exactly the same thing to illustration and design.

What Really Worries Artists — and What They Are Confusing

The article lists grievances: lack of consent, attribution, compensation, style copying. Let us examine each honestly.

Style copying is painful, but it is not new. On DeviantArt, an interesting new style gets copied the very next day — without any AI involved. Style has never been a subject of legal protection in any jurisdiction. AI has merely accelerated and scaled what was already happening.

Legal lawsuits are not built on style but on fact: specific images were physically present in the LAION-5B dataset, downloaded and used for commercial purposes without the authors' permission. This is closer to a real violation — but even here the boundary is blurred. If you train a model on photographs of interiors where paintings hang on walls, purchased by the homeowners — who is the infringer? The law does not yet have an answer.

The demand for attribution from AI seems strange when we never demanded it from human artists inspired by each other's work. This is not an argument — it is a symptom of disorientation.

Compensation is the only genuinely strong argument. Companies earned billions not from specific images but from models for whose creation those images were indispensable. The Spotify analogy works here: the platform pays authors not because it reproduces a specific track, but because it uses the entire catalogue as the foundation of its business. The logic of Getty Images and Sweden's STIM is exactly this — and it is convincing.

But even winning every lawsuit and receiving royalties will not bring back the corporate illustration market. It is gone — just as photographers lost the family portrait market.

Where the Real Protection Lies

Practice shows that those who survive are not those who litigate, but those who keep moving.

Unique technique is real protection. An authorial system built on strict geometric curves, non-trivial mathematical foundations, rare combinations — this is opaque to AI. It averages what appeared frequently in the dataset. What is rare and original it cannot reproduce — it struggles even to describe it.

Fusion — combining two or more styles in non-trivial combinations — produces works that AI cannot decompose into base layers if that particular combination was absent from its training. Papercut art plus liquid art: AI sees the result but cannot see the structure.

However, there is a crucial practical nuance here. "Idea → result" is not yet a straight arrow. AI in its current mass form does not know your visual language. It averages. It drifts. It draws things you did not ask for. Ask it for a shadow from a fountain — and it will try to draw the fountain too. This is not a flaw that will be patched in the next update. It is a fundamental property of a model trained on averaged mass data: it does not understand local logic, a part without the whole.

This means the gap between intention and realisation is still real. And as long as that gap exists, craft does not disappear — it simply changes its form.

Selling process, not just result — the most sustainable model. An artist who sells not only paintings but brush profiles, techniques, and tools sells something AI does not produce. AI uses brushes but does not sell them.

Live contact with the audience creates attachment to the author as a person, not to the genre. This is what economists call switching cost — viewers may go to AI for an image, but they come back to the artist for the human being.

The Deficit That Was Always There

The deepest problem is neither technical nor legal. An idea is never protected by copyright anywhere — only its specific realisation. DeviantArt demonstrates this daily: a new idea survives one day before the first imitator appears. AI merely accelerates an already existing process.

The real deficit — of original ideas — existed long before generative AI. Where did the calligraphers go? They still exist, but the bulk of calligraphy is now produced by electronic and software means. This is not a tragedy — it is a civilisational shift.

What Comes Next — and Why It Cannot Be Stopped Either

There is one more dimension that changes the entire picture.

Right now AI is a mass tool with averaged models. This is like a typewriter: everyone gets the same font. But the moment is approaching when an artist will be able to train a model on materials they have personally selected — including their own paintings, their own visual language, their own logic of form. And this cannot be stopped either.

A personally trained model closes the loop. It knows your visual language. It understands your local logic. It realises your specific intention rather than an averaged one. The gap between intention and realisation — the shadow without the fountain — becomes your problem to solve, not a limitation imposed by someone else's model.

This is no longer "a thin client replaced the specialist." This is the specialist acquiring a personal thin client. A qualitatively different situation. And it destroys the linear picture of "AI displacing the artist." The real trajectory is more complex: mass AI displaces the mass market, but personal AI amplifies the unique author.

The Real Scale of What Is Happening

The authors of articles like this one and the participants in court proceedings are discussing compensation for the past. That is understandable and humanly just. But the future is already structured differently.

The real question is not legal. It is civilisational: what to do with the mass release of creative professions — just as the industrial revolution released manual labour, and digital photography released portrait photographers.

History gives no grounds for panic — every such shift generated new niches, new forms of mastery, new markets. But it gives no grounds for illusion either. What is gone is gone forever.

The artists filing lawsuits are fighting for compensation for the past. Those who will thrive are the ones who understand that the instrument has changed, build a personal relationship with their audience, develop techniques that cannot be averaged, and — when the moment comes — train their own model on their own material.

That is not the end of craft. That is craft in its new form.

And we have to live with that.

Four Trees of Technological Progress. What Usually Goes Unseen.

oleg kholin — Mon, 23 Mar 2026 09:24:31 +0000

Everyone knows the game Civilization. It has a technology tree — you research writing, mathematics unlocks, then astronomy, then navigation. Clean and logical.
But the real technological tree is more complex. Behind every node of the visible tree hide three invisible trees. And without them the main tree simply cannot grow.
Let us go through this with concrete examples.

Tree 1 — The Main Tree of Technologies
This is what everyone sees. A new idea or scientific discovery gives birth to a new technology.
Einstein in 1917 described the theory of stimulated emission of light. In 1960 Theodore Maiman built the first working laser. In 1903 the Wright brothers understood how to control the lift of a wing. In 1938 Otto Hahn split the uranium atom. These are all nodes of the main tree — ideas that change the world.

Tree 2 — The Tree of Instruments
Every new technology requires a new instrument for its realisation. This is the first hidden tree — the direct manifestation of the main tree.
The laser required a ruby crystal of special purity, mirrors accurate to the nanometre, a flash lamp of strictly defined power. The Wright brothers' aircraft required a light petrol engine — a steam engine was too heavy. The atomic bomb required centrifuges for separating uranium isotopes — without them uranium enrichment is physically impossible. The transistor required monocrystalline silicon of extraordinary purity — 99.9999999%.
It would seem that is all. There is an idea, there is an instrument for its realisation. But this is where things get truly interesting.

Tree 3 — The Shadow Tree of Instruments
The new instrument from Tree 2 cannot be created with old instruments. To produce it, entirely new auxiliary instruments are needed — not for the final product, but specifically for the creation of the main instrument.
To build the ASML lithographic machine for chip production — special machines for polishing lenses to atomic-level precision were needed. These machines did not exist before the task arose. To build the engine for the Wright brothers' aircraft — Charles Taylor, their mechanic, constructed a special lathe for turning aluminium cylinders. The existing machines of that time did not provide the necessary precision. To build the centrifuges for uranium enrichment in the Manhattan Project — special balancing machines had to be built, because rotor imbalance at 50,000 revolutions per minute destroyed the centrifuge within seconds. To grow monocrystalline silicon for transistors — Gordon Teal at Bell Labs in 1950 built a special Czochralski apparatus with temperature control accurate to fractions of a degree. Nothing like it existed in industry at that time.

Tree 4 — The Shadow Tree of Technologies
But that is still not the bottom. To create the auxiliary instruments from Tree 3 — new technologies are needed specifically for their production. This is the fourth tree, the most invisible of all.
To polish the lenses for ASML — a new technology of ion-beam etching of glass was required, which simply did not exist before. It was developed by the company Zeiss specifically for this task. Zeiss has existed since 1846 and today is the sole manufacturer of these lenses in the world. To balance the rotors of centrifuges in the Manhattan Project — a new technology of dynamic balancing of rotating bodies at ultra-high speeds was required. Before this, balancing had only been applied to steam turbines at speeds hundreds of times lower. To control the temperature in the monocrystalline silicon growing apparatus — new thermocouple and temperature regulator technology was required, with precision unachievable by industrial equipment of that era. To build Taylor's lathe for aluminium cylinders — a new technology of aluminium cutting was required. Aluminium was then a new metal and would "smear" ordinary cutting tools — special geometry and cutting speeds were needed.

From Electronics to AI — The Same Four Trees
Let us apply the same system to artificial intelligence — one of the most complex technological chains in the history of humanity.
Tree 1 — The Main Tree of AI Technologies
In 1943 mathematician Warren McCulloch and neurophysiologist Walter Pitts described a mathematical model of a neuron — a binary element that either fires or does not. This was pure theory, with no practical application.
In 1957 Frank Rosenblatt created the perceptron — the first learning neural network. But it could only solve linearly separable problems and hit a dead end.
In 1974 Paul Werbos described the backpropagation algorithm. The idea allowed a network to learn from its errors, adjusting weights at each layer. In 1986 Geoffrey Hinton, Rumelhart and Williams republished it and made it practical. This thawed AI research after years of "winter".
In 1997 LSTMs appeared — Long Short-Term Memory networks that could work with sequences and retain context. But they processed text strictly one word at a time — parallel processing was impossible.
In 2017 researchers at Google published the paper "Attention Is All You Need". The transformer architecture made it possible to process all text simultaneously rather than sequentially — each token is compared against all others through the attention mechanism. From this grew GPT, BERT, Claude and all of modern AI.
Tree 2 — The Tree of AI Instruments
The transformer architecture requires processing enormous matrices of numbers simultaneously — billions of matrix multiplication operations per second. A CPU is physically incapable of this — it is sequential. What was needed was a GPU — a graphics processing unit, architecturally designed for parallel computation.
Training modern large language models requires not just GPUs, but thousands of GPUs working simultaneously. By 2024 Meta had 600,000 H100 GPUs dedicated to AI research and development. Elon Musk's startup xAI built the Colossus supercomputer with 200,000 NVIDIA H100/H200 GPUs.
Beyond GPUs, special chips with high-speed HBM — High Bandwidth Memory — were needed, which ordinary GPUs did not have. And specialised server platforms with hundreds of gigabytes of RAM and high-speed inter-processor NVLink connections.
Tree 3 — The Shadow Tree of AI Instruments
GPUs for AI contain billions of transistors on an area the size of a fingernail. To manufacture them, lithographic machines with extreme ultraviolet radiation are needed — EUV lithography machines from ASML. These are machines the height of a double-decker bus, costing 150 to 200 million dollars each. Before the 2010s such machines simply did not exist — they had to be created specifically for this purpose.
To manufacture GPU chips, ISO Class 1 cleanrooms are required — spaces in which no more than 10 particles of size 0.1 micron are present per cubic metre of air. This is millions of times cleaner than ordinary air. Creating such spaces required new filtration systems, new construction materials, new working protocols.
To connect thousands of GPUs together into clusters for AI training, specialised high-speed InfiniBand and NVLink switches were needed — equipment that before the AI era existed only in narrowly specialised supercomputers and was not commercially available at the required scale.
Tree 4 — The Shadow Tree of AI Technologies
To build the ASML EUV lithography machine, a new technology for generating extreme ultraviolet light was required — a laser beam strikes a droplet of tin in a vacuum 50,000 times per second, creating a plasma at the temperature of the surface of the Sun, which emits EUV light. This technology was developed from scratch over more than 20 years through the joint efforts of ASML, Zeiss and dozens of other companies.
To create HBM memory for GPUs, a new technology of three-dimensional chip packaging was needed — TSV, Through-Silicon Vias. This is a method of vertically connecting silicon layers through microscopic holes. Before the AI era, such packaging density was needed by no one.
To train transformers on thousands of GPUs simultaneously, a new GPU programming technology was needed — CUDA. It was released by NVIDIA in 2007 and became central to NVIDIA's strategy of positioning the GPU as a universal tool for scientific applications. By 2015 CUDA development was increasingly focused on accelerating machine learning workloads. NVIDIA developed specialised libraries — cuDNN for deep learning and cuBLAS for linear algebra. CUDA is not simply a library. It is an entirely new technology for programming parallel computation, which did not exist before it.

The Conclusion — The Correct Order of Growth:
A step forward in the main tree of technologies actually looks like this:
First a new idea or technology emerges. To realise it, a new main instrument is needed. But that instrument cannot be created with old instruments — new auxiliary instruments are needed. And to create the auxiliary instruments, new auxiliary technologies are needed.
That is, the real direction of movement is reverse — first Tree 4 grows, then Tree 3, then Tree 2, and only then does Tree 1 become possible.
For AI this looked as follows: first in 2007 CUDA appeared — a new GPU programming technology (Tree 4). This made possible the creation of next-generation GPU clusters — H100 and A100 (Tree 3). Those in turn provided the instruments for practically training transformers at industrial scale (Tree 2). And only then did the theoretical idea from the 2017 paper "Attention Is All You Need" become GPT-4, Claude and all of modern AI (Tree 1).
Note: the transformer idea appeared in 2017. But the infrastructure for its realisation was being built from 2006 to 2020 — in parallel and independently. No one planned this as a unified programme. It simply happened that at the right moment all four trees were sufficiently grown for AI to become possible.

Why This Matters
This is precisely why genuine technological progress is so slow and expensive. Behind every visible step stand three invisible steps that had to be taken first.
And this is precisely why when countries want to stop someone else's progress — they block not the final products, but Trees 3 and 4. They do not ban rockets or aircraft. They ban ASML lithography machines, special machine tools, precision bearings, metal alloying technologies. Whoever owns Trees 3 and 4 controls Trees 1 and 2. Always.

From Icon to Post-Reality: The Evolution of Ontological Synchronisation

oleg kholin — Sun, 22 Mar 2026 08:19:11 +0000

The history of how humans engage with reality is often described as a sequence of technologies, media, or interfaces. But a more precise perspective emerges if we focus not on carriers of information, but on mechanisms of ontological synchronisation — that is, the ways in which different subjects come to share an understanding of what counts as real, existent, and actionable.

From this angle, media cease to be mere channels of transmission. They become mechanisms for coordinating reality itself.

The Icon as a Point of Shared Attention

The primary level of synchronisation is associated with iconic forms — images that preserve a resemblance to what they denote. Their function is not explanatory but referential: they fix a point of shared attention.

An icon does not require a shared language or a shared theory of the world. It creates a minimal common ground: the ability to point to “this” as a shared focus.

However, this does not produce coinciding ontologies. Each participant retains their own interpretation. What is shared is the act of reference, not the meaning.

The Interface as Operational Synchronisation

With the development of digital environments, iconic forms cease to be rare objects and become a universal interface layer. They no longer merely indicate; they organise action.

At this point, a shift occurs from synchronisation of perception to synchronisation of operations. People begin to share not only what they see, but what they can do.

Reality in an interface environment becomes operational: it is defined not by what “is”, but by what can be interacted with.

Video as Synchronisation of Time

Video introduces a fundamentally new layer — temporal continuity. If the icon fixes an object, video fixes change.

This radically intensifies the effect of reality, because it creates the illusion of a world that exists continuously over time. Yet this continuity is not a property of the world itself, but a property of the medium.

Synchronisation here occurs not through objects and not through actions, but through the shared experience of a flow of events. People begin to share not only “what is”, but “how it unfolds”.

The Mirror and the Splitting of the Shared Object

A classical mirror introduces a fundamental rupture: the same object produces different images depending on the observer’s position.

This creates the first stable form of ontological divergence within apparent commonality. Everyone “sees the same thing”, yet each reconstructs a different spatial configuration.

A paradox emerges: commonality is preserved at the level of the object, but breaks down at the level of experiential structure.

The Algorithmic Mirror and the Personalisation of Reality

With the emergence of algorithmic systems, reflection ceases to be neutral. It becomes computed, adaptive, and predictive.

Thus appears the controlled mirror: the system does not simply display reality but constructs its representation based on a model of the user, their behaviour, and context.

This produces a shift: a shared reality fragments into personalised versions, each internally coherent and functional, but not required to coincide with others.

Virtual Reality as Coordinated Simulation

In multi-user virtual environments, it becomes necessary to maintain consistency between participants. The world is no longer given in advance; it is computed as a shared result of interaction.

However, this commonality does not imply identical perception. It implies coordinated action within a shared simulation.

Reality becomes protocol-based: it exists insofar as the system maintains non-contradictory interaction among participants.

AI as an Active Mirror and Generator of Ontology

With the introduction of AI, a qualitative shift occurs. The system ceases to be a passive reflector or a fixed simulator. It becomes an active generator of the environment.

The AI mirror does not simply show the world — it constructs it. This construction depends on the model of the subject: the system takes into account behaviour, preferences, and responses, forming an adaptive ontology.

A crucial point: a feedback loop emerges. The user does not merely perceive the world; they become an input into its generation.

Post-Reality as a Mode of Coordinated Generation

As a result, a condition emerges that can be described as post-reality.

Its key feature is not the disappearance of reality, but the disappearance of a single external ground against which “true” and “false” could be distinguished.

Reality is no longer given as an external object. It is continuously produced — through interaction, synchronisation, and computation.

The criterion shifts from correspondence to the world to the stability and operability of the generated ontology within the system of interactions.

Conclusion: From Shared Object to Distributed Ontology

The evolution of ontological synchronisation follows a последовательный shift:

from a shared object that can be pointed to
to a shared interface through which action is performed
to a shared flow of events that can be experienced
to personalised reflection that shapes experience
and further to active generation of reality adapted to interaction

In this perspective, reality ceases to be a stable stage. It becomes a process of coordination — a dynamic system in which ontologies do not simply coexist but are continuously produced and recalibrated.

Post-reality, in this sense, is not the end of the distinction between truth and illusion, but a transition to a regime in which that distinction is no longer fundamental.

Robots and Production: How Not to Freeze System Evolution

oleg kholin — Sun, 15 Mar 2026 08:08:23 +0000

We live in an era of robotic hype. Humanoids perform backflips in videos, "smart" manipulators pour water, investors pour money into promises of the future. But behind all this hides a fundamental problem that few notice.
The Investment Race, Not Engineering Necessity
The modern race for anthropomorphic robots is not about technological progress. It's about investments. Whoever promises more gets more. A video of a robot doing a backflip or "pouring water" is not an MVP (minimum viable product), but an MVD (minimum viable demonstration) for attracting capital.
Investors react to anthropomorphic gestures — a smile, a nod, "natural" movements — not to metrics of uptime or production cycle cost. The cost of developing a humanoid at $50-100 million is recouped not through sales (the industrial humanoid market is practically zero in 2024-2025), but through increased market capitalization of the parent company, grants for the "future of work," and selling a technological image.
Anthropomorphism today is not a design error, but a rational strategy in a distorted market where what's measured is not functionality, but virality.
Robot as a "Profile in Search of a Task"
But there's a deeper problem. A robot is created as a profile, but there's no narrative explaining it. And this is because there's no context for it. The robot is being forced into something, rather than building a process around the robot.
Here's an example. A robot can expertly insert rubber insulation into windows of new cars. Great! But who unpacks them and places them on the table? Another robot? Or a human?
This is a perfect illustration of the point automation syndrome. They automate the "headache" — the complex operation, ignoring the "tail" tasks: logistics, preparation, cleanup. The result is a robot-island requiring human servicing at the input and output. This is not process automation, but delegation of a single movement.
Freezing Production Dynamics
But you're missing another critically important point. Implementing robots is freezing the dynamics and mechanics of embedded production for several years!
With a human as the operational unit, the operational implementation of new materials, components, and technologies happened relatively quickly. A human operator is cognitively and operationally flexible. But with an implemented robot, new operations as mechanics can still be implemented, but how do you implement a new complex interaction of the robot with new materials or environments? This is simply a quiet horror for technologists!
A human is a universal interface to uncertainty. Their cognitive and motor plasticity allows them to absorb environmental variability without changing the "hardware." New material? A human felt it with their hands — understood how to handle it (seconds or minutes). A robot requires reprogramming, testing, calibration (days or weeks). Change in part geometry? A human adapted their grip "on the fly." A robot needs a new gripper, possibly a new manipulator. A new operation in the cycle? A human learns in 1-2 shifts. A robot needs trajectory redesign, collision checking, safety protocols.
The cost of adaptation for a human is salary and training time. For a robot — engineering hours, line downtime, possible retooling.
When you implement a robot on a production line, you're not just automating an operation — you're fixing the process ontology. Before the robot: Material → Human → Product (flexibility at the operator level). After the robot: Material → Robot version 1.0 → Product (change = project).
The consequences are catastrophic. The innovation paradox: the more you invest in automation, the more expensive any change becomes. This creates inertia against implementing improvements. An example from the auto industry: a plant fully robotized for Model X cannot quickly switch to Model Y without multi-million dollar investments in retooling. Meanwhile, an assembly line with humans adapts in weeks.
The "technological cocoon" effect: the robot becomes a hostage to its own efficiency — it's perfect for the current task, but vulnerable to future changes. Robotization is insurance against current costs at the price of losing future flexibility.
Efficiency vs. Evolution: Antagonism by Definition
Efficiency is convergence. Direction toward one optimal point, minimization of deviations and variability, specialization, standardization, optimization for current conditions. Result — peak performance in a narrow window of conditions. Analogy — a cheetah, the fastest runner, but only on flat savanna.
Evolution is divergence. Direction — expanding the space of possibilities, metric — preserving variability and ability to find new solutions, strategy — plasticity, redundancy, "backup" paths. Result — survival in uncertain and changing conditions. Analogy — a rat, not the fastest, not the strongest, but surviving everywhere.
Key insight: maximum efficiency equals maximum vulnerability to changes. This is not an opinion — it's a law from systems theory and evolutionary biology. The more a system is optimized for specific conditions, the more expensive any deviation from those conditions becomes.
The "freezing" mechanism is simple. Production with humans: Material → Human → Product, human adapts to new things in hours or days, in 6 months new materials, technologies, products — system evolution continues. Production with robots: Material → Robot version 1.0 → Product, fixed logic, trajectories, grippers, in 6 months new materials require reprogramming, STOP, project, budget, engineers, line downtime, system evolution frozen for months.
Cost of adaptation as a barrier. New supplier with different packaging: with human plus 1 day of training, with robot plus 2 weeks of project and $50,000. New material for a part: with human plus 1 shift of adaptation, with robot plus 1 month of retooling and $200,000. New operation in the cycle: with human plus 2 days of training, with robot plus 3 months of development and $500,000.
Result: every change becomes economically painful, creating inertia against innovations.
Biological Metaphor: Why Dinosaurs Went Extinct
The "efficient giant" scenario. Era of stability (100 million years): dinosaur huge, efficient, dominant, diet optimized for specific plants, environment stable climate, predictable conditions, result peak efficiency in its niche. Era of changes (meteorite): change in climate, vegetation, ecosystem, dinosaur too big, too specialized, adaptation impossible (slow reproduction, narrow diet), result extinction.
Survivors: small mammals inefficient but plastic, diet omnivorous, adaptive, environment can live in burrows, change behavior, result evolutionary success → human.
Transfer to production: dinosaur = fully robotized line for one model, mammal = hybrid system with human-adaptor, meteorite = new material, sanctions, market change, crisis.
Strategy "Stable Routine + Flexible Reserve"
But there's a way out. Let's return to the idea of atomizing the production process. Identifying rarely-changing routine — arranging boxes on shelves, size changes but shelves don't — optimizing a robot for it. In the end, we know this won't change much for another 7 years. Like in Amazon warehouses!
A robot should not serve as a red line for mass production! It's for lowering costs through increasing work volume! If at Ferrari production a robot stopping means the entire plant stops, then at a regular plant a human can also install the sealant!
This is a practical doctrine of reasonable robotization. Each operation is evaluated on two axes: stability (how often does this operation change?) and criticality (what happens if it stops?).
The decision matrix for robotization looks like this. Stable operation, non-critical (there's a reserve) — ideal zone for a robot. Example: warehouses, sorting, stacking. Stable operation, critical (stoppage = line stop) — robot plus mandatory reserve (second robot or possibility of manual work). Dynamic operation, non-critical — hybrid system, robot plus human backup. Dynamic operation, critical — human or hybrid system, a robot here creates too many risks.
Amazon example: arranging boxes on shelves — stable plus non-critical operation. Ideal zone for a robot. Even if Kiva breaks down, a human can temporarily take over the operation (slower, but without stopping). Success factors: operation stability (arranging boxes doesn't change for 10+ years), standardization (all boxes → one gripper type), scale (thousands of robots → scale effect), redundancy (if one robot breaks — neighboring one covers the zone), human factor (humans on picking and exceptions). Result: robots lower costs through volume, not through uniqueness.
Ferrari example: robot-welder in a unique line — stable plus critical operation. Problems: operation stability low (each model is unique → robot tied to specific program), standardization low (each part is unique), scale small (10-20 thousand cars per year), redundancy minimal (robot = bottleneck), human factor (humans cannot quickly replace the robot due to specialization). Result: robot improves quality, but creates a failure point. This is justified for a premium brand, but catastrophic for mass production.
Practical Algorithm for Robot Implementation
Step one: decomposition of the process into atomic operations. Example: installing a sealant in a car window. Operation 1.1 — getting the sealant from the magazine (stable). Operation 1.2 — positioning the sealant (stable). Operation 1.3 — inserting into the groove (stable). Operation 1.4 — checking the fit (dynamic — depends on the batch). Operation 1.5 — correction for deviation (dynamic — exceptions).
Step two: evaluation on two axes. Operation 1.1 — stability high, criticality low, decision robot. Operation 1.2 — stability high, criticality low, decision robot. Operation 1.3 — stability high, criticality medium, decision robot plus reserve. Operation 1.4 — stability low, criticality high, decision human. Operation 1.5 — stability low, criticality high, decision human.
Step three: designing redundancy. Stable operation with robot: robot version 1.0 works 95% of the time, upon failure or maintenance a human temporarily takes over the operation, the line continues to work (30% slower, but without stopping).
Step four: economic calculation. Total robotization: investment $5 million, efficiency plus 50%, flexibility 0, downtime risk $100,000 per hour, payback period 4 years. Selective robotization (our approach): investment $1.5 million, efficiency plus 30%, flexibility high, downtime risk $30,000 per hour, payback period 1.5 years.
Selective robotization gives a smaller peak of efficiency, but greater resilience to changes. This is not a compromise — it's a strategic choice.
Why Don't Everyone Do This? Barriers
Barrier one: investment hype. Investors want a "fully automated plant" — this sells better than "robots where it's stable." Our approach is less viral, but more sustainable.
Barrier two: hierarchy of engineering thinking. Engineers love "clean solutions" — a fully automated line looks more elegant than "robot here, human there." Our approach requires systemic thinking, not engineering perfectionism.
Barrier three: organizational inertia. After implementing a robot, management doesn't want to admit that a "Plan B" with a human is needed — this seems like a retreat. In reality, it's insurance.
Three Principles of Reasonable Robotization
Principle one: robot where it's stable. Not "where it's complex," but "where it doesn't change." Amazon warehouses work because boxes don't change. If unique shapes arrived every day — the system would collapse.
Principle two: a robot should not be a "red line." If a robot stopping equals the entire production stopping, you've created a failure point, not efficiency. There should always be Plan B — a human or backup module.
Principle three: robot equals cost reduction through volume. Not through "intelligence," not through "anthropomorphism," but through scalable routine. If a robot doesn't allow increasing volume or reducing unit cost — it doesn't pay off.
Final Formulation
Robotization is not a war between humans and machines. It's a search for optimal role distribution: the machine takes stable routine, the human preserves flexibility. The winner is not the one who automated more, but the one who preserved the ability to evolve.
We build systems for increasing efficiency, but lose the ability to evolve. We build "dinosaurs" and call it progress. But evolution rewards not the most efficient, but the most plastic.
The true challenge of future engineering is not to create the perfect robot, but to design a system that can evolve without self-destruction. Robotization is not an ideology, it's a tool. And like any tool, it should be applied where it's truly useful, not where it looks impressive.

GitHub for Civilizational Expansion: From an Antarctic Train to the Ontology of Autonomous Infrastructure

oleg kholin — Fri, 13 Mar 2026 20:44:52 +0000

I. The Problem of Environment
Antarctica is an environment of absolute extremes. Temperatures drop to −60°C and below, wind loads reach destructive levels, and the snow-ice surface conceals cracks, cavities, and zones of loose snow beneath it. Infrastructure is virtually nonexistent, and distances between points are measured in hundreds and thousands of kilometers.
Historically, mobility in these conditions has relied on two types of solutions: aviation and ground convoys of tracked tractors. Both approaches are fundamentally limited. Aviation depends on weather conditions and cannot carry heavy payloads. Tractor convoys depend on fuel, lack integrated scientific infrastructure, are vulnerable to snowdrift burial during extended stops, and require constant surface reconnaissance due to hidden crevasses.
Existing solutions have reached their ceiling. A paradigm shift is required.

II. The Modular Train as a New Paradigm
The response to the limitations of existing systems is the concept of a modular autonomous train — not a vehicle, but a mobile infrastructural platform.
The architecture rests on several principles. Each module is equipped with its own tracked undercarriage, which distributes traction and ensures fault tolerance: the failure of one module does not stop the system. Flexible couplings with longitudinal and lateral articulation allow the train to navigate uneven terrain. A system of retractable support struts with heating elements lifts the train above the surface during stops — similar to polar research buildings mounted on stilts — eliminating burial by snowdrift and freezing of the tracks to the ice.
Inside, the train accommodates living quarters, laboratory units, communication and computing centers, storage and technical sections. The result is a mobile scientific station with an integrated transport function — not a vehicle carrying scientific equipment, but a scientific platform that moves.
The key element is an energy module with high autonomy, providing continuous power to traction drives, heating systems, scientific equipment, and surface reconnaissance systems without dependence on external fuel supplies.

III. The Technological Foundation Already Exists
Approximately 85% of the necessary components exist right now — across different industries, dispersed, but in ready or easily adaptable form.
Tracked platforms for extreme conditions are manufactured by the mining industry. Cold-resistant steels rated to −60°C are produced in series. Thermal insulation solutions for polar conditions have been proven on existing stations. GPR ground-penetrating radar is used in geological prospecting. Starlink already provides communications at several Antarctic stations.
Tesla has opened patents on battery pack architecture, thermal management at low temperatures, and power electronics. SpaceX has effectively trained a generation of engineers through the transparency of its technical presentations and detailed post-mortems of every failure. The predictive control algorithms developed for Falcon 9 landings are directly applicable to platform stabilization on uneven surfaces.
The challenge is not inventing components from scratch — it is integrating what already exists. No one has done this yet.

IV. AI as the Central Nervous System
The volume of data generated by such a platform cannot be processed manually. Ground-penetrating radars produce terabytes continuously. Pressure, temperature, wind, and vibration sensors form hundreds of parallel channels.
The AI core does not merely process data — it predicts. A radar detects a crevasse 200 meters ahead, and the system calculates a route around it before the hazard becomes a threat. Analysis of ice surface patterns allows risk zones to be anticipated before they are reached.
The architecture is built on the principle of majority logic: three independent sensors operate simultaneously and vote among themselves. If two out of three agree — that is the accepted value. Not a backup sensor in case of failure, but three independent sources operating in parallel as a permanent operational norm.
A critical requirement: the system continuously explains itself to the operator even during routine operation. It does not simply act — it maintains an accurate picture of what is happening for the human in the loop. Without this, the automation paradox emerges: the better the AI performs, the less capable the operator becomes at the moment when the system does eventually make an error.

V. Antifragile Interfaces
Antifragility applied to interfaces means: the system does not merely withstand stress — it improves because of it.
The failure of a sensor unit does not cause a system collapse — the load is redistributed to adjacent nodes. The interface does not crash; it simplifies to the minimum necessary. In normal operation — a full interface with analytics. In a critical situation — automatic narrowing to three to five key parameters. Operator cognitive overload at the moment of crisis is eliminated by architecture, not by training.
Every abnormal situation improves the system's behavioral model. A platform that has survived a blizzard knows more about blizzards than it did before. The knowledge base accumulates directly in the field, not in a laboratory.

VI. The Chrome Philosophy: Changing the Paradigm
When Google launched Chrome in 2008, the browser seemed like a solved problem. The team reframed the foundational assumption: not "browser as application" — but "browser as operating system." Each tab is a separate process. The crash of one does not bring down the rest. Antifragility implemented architecturally.
The same logic applies here. Not an improvement of existing convoys — a redefinition of what polar infrastructure is. Not a vehicle carrying scientific equipment — a mobile scientific corridor generating continuous longitudinal profiles of ice cover, atmosphere, and subsurface structures across thousands of kilometers.
The development methodology follows the same philosophy: a small team with high autonomy, rapid iterations, testing to failure, analysis, next version.

VII. Abstracting Every Node
The decisive step is abstracting each functional node from the specific environment in which it operates.
What is being built is not "a radar for ice." It is "a universal forward obstacle interface." The input can be any type of surface data: GPR radar for ice, LiDAR for rocks and sand, thermal imaging for marshland, seismic sensors for unstable ground. The output is always the same: a unified traversability map. The core does not know which sensor is active. It receives the map and makes a decision.
What is being built is not "protection against −60°C." It is "thermal homeostasis of a module." The environmental parameters change — the logic of maintaining internal conditions remains universal.
What is being built is not "an energy system for Antarctic conditions." It is "an abstract power source": a nuclear reactor, a diesel generator, solar panels in the Sahara, hydrogen fuel cells — the output is always a stabilized power flow.
This is a Hardware Abstraction Layer — not for a single device, but for a physical platform in its entirety. Windows does not know what hardware is inside it — it operates through an abstract driver layer. The same principle is applied here to mobile infrastructure in any hostile environment.
Today Antarctica. Tomorrow the Sahara. The day after, the Moon. The core of the platform does not change. The adapters do.

VIII. An Ecosystem, Not a Product
The train is one element of a broader system.
The concept includes an autonomous base with an AI decision-making core, robotic manipulators, and 3D printers capable of producing structures from local materials. The train delivers data and mechanical constructions to the base. The base autonomously deploys and begins operation. Once sufficient resources have been accumulated, it deploys the next node. The train carries the seed kit for the new base.
This is a replication protocol. Self-reproducing infrastructure with minimal human involvement.
In this logic, Antarctica is not the destination — it is a proving ground with one key advantage over space: it can be reached and tested under real conditions. Every failure, every engineering decision, every adaptation is real data, not simulation.
Earth → Antarctica (proof of concept)
→ Moon
→ Mars
→ Asteroid Belt
The logic is the same as SpaceX applied: not flying to Mars immediately, but first learning to land a rocket on a barge in the ocean. The Antarctic train is that barge. The first validated node in a long chain.

IX. The GitHub Precedent
Before GitHub, code existed — but it was closed, fragmented, and incompatible across teams and organizations. GitHub did not invent git. What was created was a social infrastructure around an existing tool. The result: any team anywhere on the planet takes any code, adapts it, and returns the improvement to the system.
The same logic applies to physical platforms. A train is not being invented. A social and technical infrastructure is being created around the ontology of autonomous expansion.
A verified module enters the open standard. Another team takes it, adapts it to their environment, and returns the improvement. The community verifies it. The solution enters the main branch.
Currently, NASA is solving the problem of autonomous deployment. ESA is solving the same problem independently. Other organizations are solving the same problem, independently, expending resources in parallel. A shared platform means the problem is solved once, verified under real conditions, and made available to all subsequent projects.
The network effect operates here exactly as it does in software: more verified modules make the platform more valuable, which attracts more teams, which raises the quality of modules. The threshold beyond which major organizations can no longer ignore the standard is reached organically.

X. A Repository for Physical Infrastructure
GitHub changed the speed at which knowledge accumulates in software. Every solution found once stopped being lost — it remained in the system and became available to everyone who came after.
The same architecture is applicable to physical infrastructure for hostile environments. Every expedition. Every failure. Every engineering solution. Every adaptation of a module to new conditions — remains in the system and informs every subsequent project.
Antarctica is the first commit. Not because it is a compelling metaphor, but because it is where real verified data on the operation of autonomous self-replicating infrastructure in an extreme environment will be obtained for the first time. Everything that follows is built on that foundation.

The Economics of Limits: From Expansion to the Cycle. The Financialisation of Recycling as a New Growth Paradigm

oleg kholin — Wed, 11 Mar 2026 18:58:40 +0000

I. The End of the Frontier
The historical logic of capitalism was built on expansion. Three successive waves drove the system forward: geographical expansion opened new markets, demographic growth delivered new consumers, and technological revolutions created entirely new categories of need. Each time one wave subsided, the system found the next.
The transfer of manufacturing to Asia throughout the 1980s and 1990s was not merely about cheap labour — it was a mechanism for financing the computer revolution and the rollout of the internet. Low-cost production subsidised expensive innovation. Billions of people were absorbed into the global economy simultaneously as producers and as consumers. This is the classical mechanics of frontier expansion.
Today, that mechanism is spent. The territory has been settled, basic needs in developed economies are saturated, and demographic growth is concentrated precisely where purchasing power is weakest. The system has run into its own ceiling. And it is here that economics becomes genuinely interesting.

II. The Monetisation of Repeatability
When a market can no longer expand outward, it begins expanding inward. This does not mean creating new needs — it means fragmenting existing ones into billable microevents. Hence the logic of the subscription economy, hence hardware-as-a-service, hence the printer that locks up without a proprietary cartridge.
A kettle on a subscription model is neither absurd nor a caricature. It is the next logical step along a path already travelled by smartphones, automobiles, and software. The device becomes cheaper; control over its usage cycle remains with the manufacturer. A heating element designed to last precisely fifty boiling cycles is not a design flaw. It is a predictable cash flow engineered directly into the hardware.
The operative word here is repeatability. Not product quality, but the reproducibility of the consumption process. The system values not how good the kettle is, but how predictably it will be replaced, updated, and repurchased. This is precisely why downloadable "floral tea brewing profiles" represent not irony but a perfectly viable monetisation layer built atop a physically stagnant product. The experience economy superimposes itself upon the object economy once the object has exhausted its growth potential.
Yet this model carries its own ceiling. Total subscription dependency breeds existential resistance in the consumer. When everything becomes a rental, the question arises: what does a person actually own? This is not philosophical abstraction — it is a genuine social fracture, already visible in the growing appetite for repairable, autonomous, offline objects. The counter-movement of "Kettle 1.0" is not nostalgia; it is a rational response to fatigue with managed consumption.

III. The Asymmetry of Crisis
The conventional view of crisis as a "reset of Maslow's hierarchy" is an appealing but imprecise metaphor. War, epidemic, and imperial collapse have historically zeroed out the level of needs and launched new cycles of growth. Schumpeter's creative destruction functioned in a world of relatively low systemic interconnection.
Today, crisis does not reset the system uniformly. It stratifies it.
A man drinking coffee on the Cannes waterfront, watching the glow of fighting above the Libyan horizon over the Mediterranean, is not a metaphor for cynicism. It is an accurate description of the architecture of modern risk. The periphery absorbs the blow. The centre continues to consume. Capital flows toward safe jurisdictions. Destruction does not destroy the system — it redistributes its resources.
At the same time, crisis generates differentiated demand for security: the poor purchase window bars and reinforced doors, the middle classes expand their insurance coverage, and the wealthy invest in private security and asset diversification. Demand does not disappear — it transforms according to purchasing power. The security market grows precisely when the threat grows.
The Yugoslav example illustrates this mechanism in its most extreme form: the destruction of legitimate distribution channels immediately spawns shadow markets where pricing is determined by scarcity rather than production. Donor organs, medicines, foodstuffs — all become "war assets" wherever the state ceases to control the market. The economic boundary of permissibility runs precisely where resources cease to be universally accessible and become instruments of power.

IV. Geopolitics as Logistics
Strikes against Iran, the new trade corridor from India, proxy conflicts in Pakistan and Afghanistan — all of this reads not as ideological confrontation but as a struggle for control over supply chains. The global system orients itself around macro-resources and logistics; local human consequences are classified as collateral damage, unaccounted for in the market price of the primary asset.
The new India–Middle East–Europe trade route is not simply an alternative logistics arrangement. It is an attempt to escape direct confrontation — to establish a corridor independent of instability zones, to secure supply predictability without betting on high-risk regions. This is precisely why anything threatening this corridor comes under pressure: not for ideological reasons, but for purely economic ones.
China understands this. Its response is unpredictable for precisely that reason — it recognises that the new route is not merely a competitor's logistics play, but an attempt to architect global trade either without China or alongside it on unfavourable terms. The India-Pakistan escalation, with Chinese air defence systems deployed on Pakistan's behalf, is no longer a regional conflict. It is a signal marking where the real lines of strategic interest lie.
Yet it is here that Western strategy carries an internal contradiction. Simultaneously financing new corridors and sustaining military pressure is a race that exhausts faster than it pays back. Competition with China at a global scale does not kill through a single blow, but through the accumulated burden on public finances, defence budgets, and technology investment. Every new route requires security guarantees. Every security guarantee requires resources. The circle closes upon itself.

V. The Financialisation of Recycling as a Response to the Limit
If expansion is no longer possible, and crisis merely redistributes rather than resets — where is the next source of growth to be found?
The answer that emerges from this line of reasoning sounds counterintuitive: in recycling. Not in the sense of waste processing as an environmental project, but in the financialisation of an object's entire lifecycle from beginning to end.
The IBM PC example is instructive here. The chassis and motherboard remain constant — everything else changes. This is not nostalgia for modular architecture. It is the description of an economic model in which value is created not through the production of the new, but through the controlled replacement of components at a predictable rhythm. Each replacement is a transaction. Each transaction is revenue. The entire lifecycle of an object becomes a financial instrument.
This is fundamentally distinct from mere "quality." Repeatability outweighs perfection. A system capable of reproducing the replacement cycle with predictable precision is worth more than one that produces a flawless product a single time. This is precisely why the financialisation of recycling is not an environmental agenda. It is the next growth model in a world of saturated markets.
The transfer of manufacturing to Asia financed the computer revolution. The financialisation of recycling may finance the next one — whatever form it takes. Closed-loop cycles, modular products, subscriptions for physical objects, digital lifecycle management of individual components — all of this already exists in fragments. The task is to assemble it into a coherent systemic model.
China, still thinking in terms of linear export, risks falling behind here. The advantage will belong to those who first build the infrastructure of closed cycles with financial control at every node — not as a by-product of environmental policy, but as the primary business model.

VI. Conclusion: Economics Without a Frontier
We are moving into a world where growth no longer means "more." It means "deeper," "more precise," "more repeatable." Insurance transforms into predictive risk management. Manufacturing transforms into lifecycle service. War transforms into logistical pressure. Crisis transforms into a permanent background condition against which some monetise anxiety, others monetise security, and others still monetise the wreckage.
This is neither pessimism nor an apologia for cynicism. It is a description of a system that has exhausted extensive growth and is compelled to invent intensive growth instead. History demonstrates that such transitions are painful, uneven, and invariably produce new winners — those who understood earliest in which direction value now lies.
Today, that direction points inward along the cycle, not beyond its edges.

From Image to Knowledge: How Images Could Synchronise Practices in Early Human Societies

oleg kholin — Wed, 11 Mar 2026 09:24:51 +0000

When we speak about the origin of human communication, discussion most often focuses on language: when it appeared, how syntax developed, and what biological changes enabled humans to speak. However, there is another way to approach the problem — by examining it from the perspective of signs and images, rather than words.
It is possible that before complex speech humans already possessed rather effective means of synchronising knowledge. And a key role in this process may have been played by iconic images — recognisable representations of important objects. But to understand why images became necessary, we must first look at how early human groups communicated without them.

The Triad of the Sign
One of the most influential models of the sign was proposed by the American philosopher Charles Sanders Peirce.
According to Peirce, any sign includes three elements:

object — that to which the sign refers
representamen — the form of the sign (image, word, symbol)
interpretant — the meaning that arises in the observer

Peirce also distinguished three types of signs:

icon — a sign resembling its object
index — a sign causally connected with its object
symbol — a conventional sign (for example, a word)

Icons are particularly important for our purposes. They are understandable even without a shared language because they preserve a resemblance to what they represent. This property becomes critical precisely when people from different groups — with different languages and different traditions — need to communicate.

Communication Within the Group: Why Language Was Not Needed
Within small, tightly organised groups, communication could function remarkably well without developed speech. The reason lies in two structural features of early societies: rigid hierarchy and shared practice.
In a group with a clear hierarchy, most coordination does not require verbal explanation. Roles are fixed, sequences of action are known, and authority is unambiguous. In such conditions, short signals — a gesture, a sound, a glance — are sufficient because the context is already shared by all participants. The situation resembles a special operations unit working in silence: each member knows their role, and communication is reduced to minimal, precise signals. Lengthy verbal explanation would not only be unnecessary but potentially dangerous.
This internal communication system relied on gesture and body language, guttural sounds and vocal signals, demonstration of actions, and joint practice. When people constantly participate in the same activity — hunting, for example — they do not need to explain every detail in words. The knowledge is embedded in the shared situation itself.
However, the situation changes fundamentally when groups encounter one another.

The Boundary Encounter: Where Language Begins
The most important — and most overlooked — trigger for the development of speech may not have been internal group complexity, but the encounter between strangers.
When members of two different groups meet, the entire internal system of communication breaks down. There is no shared hierarchy. There are no common signals. The instinctive response — fight or flee — resolves the encounter but creates no new social bond. Yet sometimes neither fighting nor fleeing was the optimal strategy. Resources could be shared. Knowledge could be exchanged. Alliances could be formed.
In this situation, something new was required: the spontaneous construction of a social bond with a stranger. And this is precisely where a shared visual reference point becomes invaluable. If both groups recognise the same animal — a bull, a bison — then pointing to an image of that animal creates an immediate common ground. It is not yet language. It is not yet negotiation. But it is the beginning of shared attention, which is the foundation on which communication can be built.
In this sense, language may have emerged not from within the group, but at its boundary — as a response to the challenge of creating social connection with the unknown.

The Cognitive Centre: the Animal
For many prehistoric societies the key objects were large animals: bulls, bison, boars, deer. They played a central role in economy, mythology, and ritual life. From a cognitive perspective such animals possess several important properties: they are visually recognisable, they are emotionally significant, and they are connected with vital practices. For this reason the image of an animal can become the centre of a cognitive network, around which different types of knowledge converge: appearance, characteristic sounds, danger, modes of interaction, ritual meanings.
But there is a further dimension. The same animal could occupy very different positions in different ontologies — different frameworks of understanding the world and one's place in it. For a hunter, the bull is prey: something to be tracked, ambushed, killed. For a herder, the bull is a managed resource — a source of labour and wealth. These are not merely different practices. They are different relationships to the world, with different values, different risks, and different knowledge systems.
Yet both groups are talking about the same animal.
This overlap — one object, two ontologies — is precisely what makes the iconic image of the bull so powerful as a point of synchronisation. It does not require either group to abandon its framework. It simply provides a shared visual anchor around which different types of knowledge can be brought into contact.

Knowledge and Experience
Here it is important to distinguish two types of knowledge.
Object knowledge is knowledge about the object itself: what the animal looks like, what sounds it produces, how dangerous it is. Such knowledge is primarily connected with recognition.
Procedural knowledge is knowledge about actions: how to hunt, how to approach, which tools to use, how to herd. It is formed through practice and experience.
Within a single community both types of knowledge usually develop together. People learn to recognise the animal while simultaneously mastering ways of interacting with it. However, in a situation of inter-group contact the order is necessarily different.
First comes shared knowledge of the object — the name, the image, the recognition. Only afterwards does the coordination of practices become possible.
This sequence is not arbitrary. Naming the object — or pointing to its image — creates a shared topic. It directs attention. It establishes the common ground from which exchange can begin. The procedural knowledge, which is far more complex and requires demonstration and practice, can only be transmitted once this common ground exists. In this sense, the image of the bull precedes the technique of hunting or herding it.

Iconic Images and Göbekli Tepe
Archaeology offers remarkable evidence for this model. One of the most significant examples is Göbekli Tepe, a monumental complex in southeastern Anatolia dating to approximately the 10th–9th millennia BCE. The site consists of circles of massive T-shaped stone pillars bearing detailed reliefs of animals: bulls, boars, foxes, snakes, birds. The images are carefully executed and easily recognisable — semiotically, they are icons, functioning through their resemblance to their objects.
The standard interpretations — ritual centre, astronomical observatory, early temple — are not wrong, but they may be incomplete.
Consider the historical moment. Göbekli Tepe was constructed at precisely the period of transition from hunting to early herding and agriculture in the Fertile Crescent. This was a period of contact between groups with fundamentally different ontologies: those who hunted wild animals and those who were beginning to manage them.
The bull appears repeatedly across the site. For a hunter arriving at this place, the image of the bull represents a prey animal — a source of food and danger. For an early herder, the same image represents a managed resource — a source of labour and wealth. The image does not resolve this difference. But it creates a shared point of focus around which both groups can orient themselves, exchange knowledge, and begin to coordinate.
In this reading, Göbekli Tepe is not simply a ritual centre. It is a node of ontological synchronisation — a place where different ways of knowing and interacting with the world were brought into contact through shared iconic images.

Ritual as a Mechanism of Synchronisation
An image alone, however, is not sufficient. For a sign to become part of social practice it must be interpreted collectively. At this point ritual appears.
Ritual performs several functions: it fixes the meaning of the sign, it synchronises the actions of participants, it establishes shared rules, and it creates emotional investment in the shared reference point. The sequence can be described as follows:
image → collective attention → ritual → coordination of practices
In this sense the image of an animal is not merely a decorative element but a point of concentration for collective knowledge — the fixed centre around which different practitioners can gather, compare their understanding, and begin the process of coordination.

Cave Paintings as Pedagogy, Not Art
As groups grew in size, a new problem emerged: how to transmit knowledge to those who did not participate in the event directly.
Joint practice remains the most important method of learning. Yet it has an inherent limitation: not everyone can be present. A hunt involves a limited number of participants. Complex strategies cannot be rehearsed in real time with the full group. Young members, those recovering from injury, and those with other roles cannot always learn by direct participation.
This is where images perform a function that has been systematically underestimated in archaeological interpretation.
Cave paintings and rock art should not be understood primarily as artistic expression or as evidence of symbolic consciousness in the abstract. They are more precisely understood as representational tools for the discussion and transmission of collective experience — diagrams of shared action, made for those who were not present.
Consider the characteristics of many prehistoric cave images: animals in motion, multiple figures in relation to one another, repeated standardised scenes. These are not portraits. They are scenarios. They show what happened, who did what, and in what sequence. They are, in essence, the earliest known form of instructional representation.
Around such images, a group could gather and reconstruct an event: discuss what went well, explain the roles of different participants, train newcomers in the logic of collective action. The image becomes a tool for the collective processing and transmission of experience.

The Elders: The First Institution of Verbal Knowledge Transfer
The model of learning through joint practice has one further limitation that is rarely discussed in the context of language origins.
Adult members of a group — the hunters, the herders, the craftspeople — are occupied. Their knowledge is embedded in action. They transmit it through demonstration, through presence, through shared work. This system is highly effective, but it depends on co-presence and physical capacity.
The emergence of care for elderly and disabled members of the group — itself a significant marker of social development — created an unexpected cognitive resource. Individuals who could no longer participate directly in hunting or herding were nevertheless repositories of accumulated knowledge. They had participated in more events, observed more outcomes, and accumulated a broader picture of what worked and what did not. Freed from the demands of physical activity, they could concentrate on the transmission of knowledge through a different channel: verbal description, narrative, and the use of images as aids to explanation.
This institutional role — the elder as teacher — may have been one of the primary drivers of the rapid development of verbal language. The capacity for verbal transmission that began as a supplement to demonstration gradually became an independent channel, capable of conveying complex knowledge to those who had never participated in the relevant practice at all.
Children, in particular, began to acquire knowledge verbally at an increasingly early age — knowledge of animals, dangers, techniques, and social rules — long before they were capable of participating in the activities themselves. The image of the bull, explained verbally by an elder, became the child's first cognitive map of a world they had not yet encountered directly.

Images as Nodes of Synchronisation
Bringing all these elements together, the following model can be proposed.
Key objects of the surrounding world — large animals above all — become cognitive centres around which knowledge is organised. Iconic images of these objects perform several interconnected functions:

Within the group: they serve as stable reference points for the transmission of experience to those who did not participate directly — through discussion, narrative, and ritual rehearsal.
At the boundary between groups: they provide a shared visual anchor that allows strangers to establish common ground without a shared language — the first step toward communication and coordination.
Across ontological differences: a single image can serve as a point of contact between groups with fundamentally different practices and worldviews, creating the possibility of knowledge exchange without requiring either party to abandon its framework.
In the development of language itself: the naming of the iconic object — the bull — provides the first shared lexical item around which verbal communication between strangers can be organised, preceding and enabling the transmission of procedural knowledge.

In this sense, iconic images may function as nodes synchronising collective experience across the boundaries of group, practice, and ontology.

Hypothesis and Its Limitations
It is important to acknowledge what this model cannot prove.
Archaeology rarely allows communication processes to be observed directly. We see stone pillars, cave paintings, and tools — but not the conversations that surrounded them. The attribution of pedagogical or synchronising functions to specific images requires inference, and inference can be wrong.
Several elements of this model are particularly difficult to verify archaeologically: the internal communication dynamics of early groups, the specific role of elderly individuals in knowledge transmission, and the sequence in which object knowledge and procedural knowledge were transmitted.
Nevertheless, the model is not purely speculative. It makes predictions that can be tested against the archaeological and ethnographic record: the standardisation of animal imagery in places of inter-group contact; the presence of sequential or instructional scene compositions; the spatial organisation of sites in ways consistent with collective gathering and discussion; the correlation between evidence of inter-group exchange and the complexity of shared symbolic systems.

The Cathedral and the Book: A Recurring Pattern
The model proposed in this article is not confined to prehistory. It describes a pattern that recurs at every major turning point in the history of human communication.
In the nineteenth century, Victor Hugo articulated an intuition that reaches far beyond its immediate context. In Notre-Dame de Paris, he proposed that the printed book would kill the cathedral — ceci tuera cela. On the surface, this reads as a claim about competing media. But at a deeper level, Hugo was describing exactly the dynamic this article has traced.
The medieval cathedral was not simply a building, and not simply a religious institution. It was a material node of ontological synchronisation — a place where people arriving with fundamentally different frameworks of experience (the peasant, the knight, the merchant, the monk, the pilgrim from a distant region) encountered a shared system of iconic images. The portals, the stained glass, the sculptural programmes functioned precisely as Peirce's icons: recognisable, visually immediate, requiring no shared verbal language to begin the work of creating common ground. You did not need to be literate. You did not need to speak the same dialect. The image of the bull on the pillar at Göbekli Tepe and the image of Christ in the tympanum at Chartres are separated by ten thousand years, but they perform the same cognitive and social function: they are points where different ontologies can meet.
What Hugo sensed was that the printed book does not simply replace the cathedral as a carrier of information. It transforms the entire structure of synchronisation. The cathedral operates through collective, embodied, spatial experience — people gather in one place, interpret images together, participate in ritual that fixes shared meaning. The book operates through individual, verbal, abstract experience — it is portable, it can be read alone, it does not require co-presence. The gain in precision and reach is enormous. But something is lost: the specific power of the shared iconic image in shared space to bring different ontologies into contact without requiring them to first agree on words.
This is not a unique transition. It is a recurring one. The pattern this article has traced — from gesture and guttural signal, to iconic image, to ritual synchronisation, to verbal transmission, to institutionalised language — does not end in prehistory. It replays at every moment when a new medium emerges and the existing nodes of synchronisation are challenged:
cave → temple → cathedral → printed book → internet
Each transition shifts the balance between the iconic and the symbolic, between collective and individual interpretation, between presence and abstraction. Each time, the new medium offers greater reach and precision. Each time, something of the older synchronising function is lost and must be reconstructed in a new form.
We are living through such a transition now. The question of how different ontologies — different practices, different communities, different ways of understanding the world — find shared points of recognition is not an archaeological question. It is an urgent contemporary one.

Conclusion
Human communication probably developed not only through words. Before the emergence of complex linguistic systems, people used gestures, vocal signals, demonstration, ritual, and iconic images as a coordinated system for managing shared knowledge.
Within this system, images of important objects — above all large animals — could function as shared cognitive anchors. They served the transmission of experience within the group, the creation of common ground between groups, and the bridging of different ontological frameworks.
The development of language itself may have been driven not by internal group complexity alone, but by the encounter with the stranger — the moment when the familiar system of hierarchy and shared signal broke down, and a new kind of social bond had to be constructed from nothing, around a shared point of recognition.
The ancient images of animals — from cave paintings to the reliefs of Göbekli Tepe, from the portals of medieval cathedrals to the icons of the digital screen — were not the beginning of art. They were the beginning of the infrastructure through which human beings learned to think together across difference. That infrastructure has changed its form many times. The chain runs from the bull carved in stone ten thousand years ago to the cathedral Hugo mourned, and from there to every new medium that promises to connect us. Each link in that chain is built on the same foundation: a shared image, a moment of mutual recognition, and the possibility of a conversation that could not otherwise begin.

Mobile Ground Train for Antarctica: Concept of an Energy-Redundant Autonomous Platform

oleg kholin — Tue, 10 Mar 2026 09:59:29 +0000

Part I. Antarctica as an Engineering Environment: Conditions and Mobility Challenges

Antarctica represents a unique natural system combining extreme climatic conditions, vast distances, and almost complete absence of infrastructure. Temperatures in the central regions of the continent can drop to −60 °C and below, wind loads reach significant values, and the snow-ice cover conceals a complex subsurface structure, including cracks, cavities, and zones of loose snow.

Historically, mobility in these conditions has relied on two main types of transport: aviation and ground vehicles. Aviation provides high speed and rapid access to remote locations; however, its application is limited by weather conditions, the availability of prepared landing areas, and comparatively low payload capacity.

Ground systems are represented by tracked tractors and sled trains used to supply polar stations and conduct expeditions. Typically, such systems are convoys of heavy machinery moving along pre-surveyed routes. Despite their logistical effectiveness, they have fundamental limitations:

dependence on fuel and limited energy autonomy;

lack of integrated observation and data processing infrastructure;

high vulnerability to snowdrifts during prolonged stops;

the need for constant surface reconnaissance due to hidden ice cracks.

These limitations motivate the search for alternative engineering solutions capable of combining transport, energy, and scientific infrastructure into a single autonomous system. One possible approach is the concept of a ground modular train—a large multi-sectional platform designed for long autonomous expeditions across the Antarctic ice sheet.

Part II. Antarctic Train Architecture

The proposed system consists of multiple specialized modules (cars) connected by flexible couplings with a certain degree of longitudinal and lateral articulation. The geometry of the cars features increased width and reduced height, which lowers aerodynamic drag in crosswinds and reduces the center of gravity of the entire system.

Each module is equipped with its own tracked running gear. Configurations with one or two pairs of tracks per car are possible, significantly reducing ground pressure on snow and increasing stability on weak surfaces. This architecture ensures distributed traction and increased fault tolerance: in the event of partial failure of a module, the train remains capable of movement.

Energy Module

A key element of the concept is a dedicated energy section housing a compact nuclear fusion energy module—provisionally called a “fusion battery.”

Such a power source, with extremely high energy density and virtually unlimited operational resource, allows:

continuous power supply to traction drives;

heating of living and technical modules;

operation of scientific equipment;

operation of active surface reconnaissance systems;

energy-intensive operations such as driving support struts.

The presence of this power source elevates the system from a fuel-dependent vehicle to an energy-redundant autonomous platform.

Surface Reconnaissance System

Traveling across the ice sheet requires continuous monitoring of subsurface structures. For this purpose, a sensor complex is installed in the front of the train, including radar systems for subsurface sounding. Radar allows detection of cracks, cavities, and weak snow layers at depths of tens of centimeters or meters ahead of the moving train.

The collected data are processed in real time, enabling route adjustments and enhancing movement safety.

Parking and Train Lifting System

One of the key challenges of Antarctic expeditions is overnight stops and prolonged parking. During storms, stationary vehicles can quickly become buried in snow, which blocks the running gear and requires laborious clearing.

To address this, a system of retractable support struts is proposed. In transit, the struts are stowed along the body and do not interfere with movement. When parked, the struts are extended and screwed into the snow and ice like pile foundations. Local heating elements in the tips are used to facilitate insertion.

Once secured, the train is elevated above the snow surface, similar to polar research buildings mounted on stilts. This significantly reduces the risk of snow accumulation around the body and eases subsequent departure. Additionally, during stops, tracks can freeze into the ice, adding another factor for the lifting system to address.

Before resuming movement, the struts are heated to a higher temperature to loosen their grip on the ice. They are then unscrewed and returned to the transit position.

Living and Scientific Modules

The internal space of the train includes multiple functional zones:

crew living quarters;

laboratory units;

communication and computing centers;

storage and technical sections.

This architecture allows the train to function not only as a transport vehicle but also as a full-fledged mobile scientific station.

Part III. Systemic Difference from Tracked Vehicle Convoys

At first glance, the concept may resemble traditional convoys of tracked tractors used to transport cargo between Antarctic stations. However, the difference between these systems is fundamental.

A tracked vehicle convoy consists of independent units, each with its own energy system, limited operational range, and a restricted set of functions.

In contrast, the Antarctic train is an integrated infrastructure platform. Its modules are connected by a shared energy system, a unified information architecture, and functional specialization. Such a system can perform tasks far beyond simple cargo transport.

Historically, attempts to create large polar vehicles have been made repeatedly. One of the most famous projects was the Antarctic Snow Cruiser, a large expedition vehicle built in the late 1930s. Despite the project’s ambition, the technology of the time proved insufficient for successful operation.

Modern Antarctic operations also rely on heavy tracked systems and sled trains, for example, at the Amundsen–Scott South Pole Station. However, these systems primarily serve logistical purposes.

The proposed concept opens new prospects. An energy-redundant mobile platform can provide continuous scientific monitoring. Traveling along extensive routes, the train can create longitudinal profiles of ice cover, atmospheric conditions, and subsurface structures.

Effectively, this is a new type of research infrastructure—a mobile scientific corridor. Instead of stationary stations spread across the continent, a dynamic monitoring system can traverse thousands of kilometers and conduct measurements along its route.

Thus, the Antarctic train should be seen not merely as a transport vehicle, but as a new form of scientific infrastructure: autonomous, mobile, and energy-independent, designed to explore one of the most remote regions of the planet.