How Markets Know What No Participant Knows
“The whole is not only more than but very different from the sum of its parts.” - Aristotle
No ant knows the architecture of the colony. No single neuron understands the thought it helps to form. No bee comprehends the geometry of the hive, the efficiency of the waggle dance, or the collective calculus that determines when a swarm abandons one home for another. And yet the colony builds. The brain thinks. The hive decides. Intelligence emerges at a level that none of its components can perceive, let alone direct.
Markets do exactly the same thing. And this is a problem for the embedded agent that is far stranger and more consequential than it first appears.
Because the market is not merely smarter than you. It is computing something you cannot access, using information you contributed but cannot retrieve, producing outcomes no participant intended and no model predicted. You are inside a distributed intelligence that includes your participation as raw material, and you have no window into the computation itself.
This is not the efficient market hypothesis. It is not the claim that prices are “correct” or that markets are rational or that beating the market is impossible. It is something more unsettling and more precise: the market computes answers that no individual computed, using processes that no individual controls, and the results of that computation are fundamentally unavailable to anyone operating from inside the system.
The Ant and the Architect
A leaf-cutter ant colony can contain millions of individuals. The colony farms fungus, manages waste, regulates temperature, defends territory, and adjusts its labour allocation in real time based on environmental conditions. It is, by any reasonable measure, an intelligent system. It solves complex optimisation problems, adapts to changing circumstances, and sustains itself across decades.
No ant knows any of this. Each individual ant follows simple rules: if you encounter a pheromone trail of a certain concentration, follow it; if you encounter a dead nestmate, carry it to the waste pile; if the fungus garden needs tending, tend it. The rules are local. The ant’s world is a few centimetres of tunnel and the chemical signals of its nearest neighbours. The colony-level intelligence, the farming, the architecture, the strategic adaptation, exists nowhere within any individual ant. It is a property of the interactions between ants, not a property of ants themselves.
This is emergence in its purest form. The colony-level behaviour is not programmed into any individual. It is not coordinated by a leader or a blueprint. It arises from the density and structure of local interactions, and it cannot be predicted by studying any individual in isolation. You could dissect every ant in the colony, catalogue every neural pathway, sequence every pheromone receptor, and you would still not find the architecture of the colony, because the architecture does not live inside any ant. It lives in the space between them.
Markets as Distributed Computation
A market is a computational system. This sounds reductive, but it is not. It is precise. Every trade is an exchange of information. Every price is the output of a process that integrates the beliefs, constraints, fears, models, errors, and strategies of millions of participants into a single number that none of those participants individually calculated.
Consider what a price actually encodes. It reflects the current inventory of market makers. The positioning of institutional portfolios. The trigger levels of systematic strategies. The emotional state of retail participants. The hedging requirements of options dealers. The rebalancing schedules of pension funds. The capital flows driven by geopolitics, by central bank policy, by the weather in agricultural regions. No single participant has access to more than a sliver of this information. Yet the price somehow reflects all of it, imperfectly, dynamically, continuously.
This is not the invisible hand. Adam Smith’s metaphor implies benign guidance, as though the market gently steers toward an optimal outcome. What actually happens is messier and more interesting. The market does not optimise. It computes. The distinction matters enormously.
An optimiser finds the best solution. A computer processes inputs and produces outputs, and the outputs are not “best” in any normative sense. They are simply the results of the computation. Markets compute prices that reflect the aggregated pressure of all participants, and those prices can be wildly wrong, persistently biased, and structurally fragile. The computation is real. Its results are not necessarily right. They are simply the emergent output of a process no one controls.
The Intelligence You Cannot Access
Here is where the embedded agent encounters a genuinely disorienting problem. You contribute to the market’s computation. Every order you place, every position you hold, every trade you execute becomes an input to the distributed calculation. Your information, filtered through your strategy, becomes part of the price.
But you cannot access the computation’s output in any useful way. The price is visible, of course. You can read it on a screen. But the price as a computational output is not the same as the price as a number. The number is trivial. The computation that produced it is opaque, and it is the computation that contains the intelligence.
Think of it this way. You can see the result of the colony’s decision. The ants move to a new nest site. But the decision process itself, the way millions of local interactions resolved into a collective choice, is invisible. You see the outcome. You do not see the logic, because the logic does not exist in any single location. It is distributed across the entire network of interactions.
Market prices are the same. The price of crude oil today reflects a computation that integrated information about Middle Eastern geopolitics, Chinese industrial demand, US shale production economics, speculative positioning, hedging flows, storage constraints, weather patterns, and thousands of other inputs. You can see the result: the price. But the process that produced it, the way all those inputs were weighted, combined, and resolved, is fundamentally inaccessible to you. Not because you lack data or intelligence, but because the computation is not happening in any location you can observe. It is happening in the space between all the participants, including you.
Smarter Than You, But Not Wise
There is a temptation here to deify the market. If it computes at a level no individual can access, perhaps we should trust it. Perhaps prices are, if not correct, at least the best available estimate. Perhaps the embedded agent’s job is simply to listen to the market’s intelligence and follow its lead.
This temptation should be resisted, because the market’s intelligence is not wisdom. The market computes without intending. It processes without purpose. It resolves information into prices without any mechanism for evaluating whether those prices serve any useful function. It is a computational engine running continuously, producing outputs, but it has no objective function. It is not trying to get prices right. It is not trying to do anything. It simply runs.
This is the crucial difference between emergent intelligence and designed intelligence. A designed system (a computer programme, an engineering blueprint) has a purpose encoded by its creator. An emergent system has no creator and no purpose. It simply does what the interactions produce. The colony does not “try” to farm fungus efficiently. Efficiency emerges from selection pressure over evolutionary time. The brain does not “try” to think. Thought emerges from the electrochemical interactions of neurons that have no concept of thought.
Markets do not try to price assets correctly. Pricing emerges from the interactions of participants who are trying to make money, hedge risk, express views, rebalance portfolios, and manage liabilities. The aggregate output of these individual motivations sometimes looks like intelligence. Sometimes it looks like madness. It is neither. It is computation, and computation without purpose can produce both order and catastrophe with equal facility.
The embedded agent sits inside this system with a peculiar awareness. You know the system computes at a level you cannot access. You know its outputs sometimes contain information that exceeds any individual’s knowledge. And you also know the system has no interest in your wellbeing, no mechanism for self-correction beyond the blunt feedback of profit and loss, and no guarantee that its computations will not produce outcomes that destroy the participants who fed them.
Why Data Will Not Save You
The natural response to distributed intelligence is to try to capture it. If the market knows something you do not, perhaps more data will reveal what it knows. Perhaps machine learning, with enough inputs and enough processing power, can crack the code of the market’s emergent computation.
This is a seductive idea and a structural impossibility. Not because the technology is inadequate, but because the thing you are trying to capture does not exist in the data. The market’s intelligence is not encoded in the data the market produces. It is encoded in the process by which the data is produced, and that process is the real-time interaction of millions of adaptive agents, including you.
Data is a record. The market’s intelligence is a process. You cannot capture a process by recording its outputs any more than you can understand a conversation by transcribing only one side. The other side is happening simultaneously, adapting in real time, responding to what you just said, and you are part of the conversation. The data shows you what the market did. It does not show you why, because the why is distributed across every participant who was active at that moment.
More data does not solve this problem. Neither does more computing power. Neither does more sophisticated modelling. These tools can identify patterns in the outputs. They can detect regularities in the record. But they cannot access the distributed computation itself, because that computation exists only in the real-time interaction of agents, and the moment you try to measure it, you alter it with your measurement.
This is the intelligence that no one has. Not because no one is smart enough, but because it is structurally unavailable to any participant operating from within the system. You are inside the brain, looking at individual neurons firing, and you cannot see the thought.
What the Embedded Agent Does Instead
If you cannot access the market’s intelligence, and you cannot capture it with data, and you cannot replicate it with models, then what is left?
What is left is a different kind of relationship with the system. Not the relationship of the scientist who decodes, but the relationship of the surfer who reads.
A surfer does not understand the ocean. Not in any computational sense. The surfer cannot model the fluid dynamics, predict the exact shape of the next wave, or calculate the forces involved. But the surfer reads the water. Through experience, through immersion, through thousands of hours of direct physical engagement with the system, the surfer develops an intuition for what the ocean does. Not what it will do precisely, but what it tends to do, what it is capable of, how it moves.
This reading is not mystical. It is pattern recognition operating at a level below explicit analysis. The surfer’s body knows things the surfer’s mind cannot articulate. And this knowledge, embodied, experiential, provisional, is often more useful than any computational model, because it is constantly updated by direct engagement with the system it seeks to navigate.
The embedded agent relates to the market’s intelligence in a similar way. You cannot decode it. You cannot replicate it. But you can learn to read the patterns it produces. You can develop, through sustained engagement, a feel for the dynamics that the distributed computation tends to generate. Not a prediction, but an orientation. Not a map, but a sense of terrain.
This is less satisfying than a formula. It is less precise than an algorithm. It is also closer to the truth of what it means to participate in a system whose intelligence exceeds your own. You are not trying to become as smart as the system. You are trying to navigate a system that is smarter than you, with tools that are suited to the task of navigation rather than the task of comprehension.
The Humility of Participation
There is something liberating in this. Once you accept that the market’s intelligence is structurally beyond your reach, you stop trying to outsmart it. You stop looking for the hidden signal that will give you the market’s view of the future. You stop believing that enough data, enough cleverness, enough processing power will crack the code.
Instead, you do something simpler and more profound. You participate. You deploy a strategy that is aligned with the structural dynamics of the system. You size it to survive the outcomes you cannot foresee. You stay engaged across conditions you did not predict. And you accept that the market will do things you do not understand, for reasons you cannot access, producing outcomes you did not anticipate.
This is not defeatism. It is the appropriate posture of an agent embedded within a system of distributed intelligence. The ant does not need to understand the colony to contribute to it. The neuron does not need to comprehend the thought to participate in it. And the embedded agent does not need to access the market’s intelligence to navigate successfully within it.
What the embedded agent needs is something different: the discipline to stay in the system, the sizing to survive its surprises, and the humility to know that the intelligence at work in the market will always exceed the intelligence of any individual participant. Including you. Especially you.
The system is smarter than you. It is not wise. It does not care about you. And you are inside it, contributing to its computation, unable to read the results.
Act accordingly.
Richard Brennan writes on systematic trading, complex adaptive markets, and the philosophical foundations of trend following at atstradingsolutions.com. His books include The Fractals of Finance and Complex Adaptive Markets. The forthcoming Carved by Impossibility completes the trilogy.
Want the theoretical foundation for why trend following works?
The Fractals of Finance: Determinism, Adaptation and the Geometry of Markets
The book explores the full architecture of feedback, fat tails, and fractal structure in financial markets, and what it means for how we trade, invest, and understand risk.
Available now on Amazon in paperback, hardcover, and Kindle.
Want a practical field manual for trading trends and capturing outliers?
The Aussie Turtles Trend Following Guide: A Field Manual for Hunting Outliers adapts the timeless principles of the original Turtle traders into a systematic, rules-based approach for modern markets. Co-authored with Adam Havryliv.
Available now on Amazon in paperback, hardcover, and Kindle.