The great modern philosopher Mike Tyson once said, “Everybody’s got a plan until they get punched in the face.” Future historians will no doubt develop many theories about the initial impetus behind this profound insight, but Tyson might as well have been referring to the main predicament of complex systems: operating and surviving more or less intact in the maze of unpredictability. Any system can develop the perfect and highly optimized operation plan, with just-in-time this and agile that, but all beautiful flowcharts can, and usually do, crumble under the onslaught of reality. The drama of all systems, both simple and complex, is in the precarious balance between maintaining operational efficiency and adapting to unforeseen challenges that punch you in the face.
To begin with, surviving amidst flux over a long period demands constant transformation from all systems. However, a continuous and dynamic tension exists between a system’s adaptability, efficiency, and overall stability. Even for simple systems, too much adaptability sacrifices efficiency, and too much efficiency sacrifices adaptability. For example, an organization could optimize its supply chain for maximum just-in-time efficiency and shareholder value, only to have it crumble under a sudden and unpredictable change in geopolitics halfway across the globe. Who knows, an obscure rebel group might start sinking container ships. True story. It gets worse, though; this tension is far more challenging for complex systems due to their being, well, the opposite of simple.
Complex systems are composed of multiple distinct elements – for example, consider a modern army with its infantry, artillery, mechanized units, drones, air force, and so on. The interactions between these disparate components lead to results we cannot predict by simply analyzing each system element’s properties in isolation. For instance, examining a tank’s capabilities in isolation offers no clear insight into its operational effectiveness when supported by infantry, artillery, drone reconnaissance, and aviation.
When combined, these additional components of the larger system radically alter the tank’s capabilities and fundamentally change the nature of its engagement envelope and effectiveness on the battlefield. The introduction of aerial surveillance provides real-time data, artillery offers long-range support firepower, infantry occupies proximal space, and aviation brings a vertical dimension, together creating a system whose potential actions are vastly different and more complex than that of any single component.
In other words, each unique element allows a system to add complexity at that scale, interface with reality differently, and engage in unique and complex behaviors. The more complexity at different system scales, the more adaptable a system is in interfacing with reality at those scales. That said, if a system gets a punch to the face and is not adaptable to deal with its effects, it quickly experiences a cascading reduction of complexity and collapses.
All systems get punched in the face sooner or later. What happens with a system after the punch is where the fun begins; the system still has to maintain internal coherence and operational efficiency while simultaneously pivoting its operations to adapt to the novel external conditions. As Mike Tyson aptly pointed out, the old plan is invalid after the punch because reality has violently imposed itself on the system’s assumptions.
Let’s explore a principle of systemic efficiency and adaptability that accounts for the punches and what comes after them. I call it Ariadne’s string principle after the ancient Greek myth of Theseus and the Minotaur.
Ariadne’s gift
The great Daedalus, legendary builder and craftsman of the ancient world, was tasked by King Minos of Crete to construct a labyrinth so complex that escape would be impossible. Once in, you were never supposed to be able to leave. On top of that, this labyrinth was to be built as a prison for the Minotaur, a creature with a man’s body and a bull’s head. Daedalus built the unique maze, the Minotaur was locked inside it, and the king put it to grim use, imposing a tribute on defeated Athens: the sacrifice of seven young men and seven young women every seven years to the Minotaur.
When the tribute was due again, Theseus, the prince of Athens, volunteered to be one of the seven young men destined for the labyrinth, pledging to slay the monster. However, our valiant hero was not alone. The daughter of King Minos – Ariadne – who, as it happens in myths, was in love with brave Theseus, approached Daedalus, pleading with the maze builder to help the hero escape. Moved, Daedalus asked Ariadne to give our hero a ball of string. Yes.
The myth does not mention Ariadne’s initial reaction to that solution, but one can imagine it. In any case, Daedalus explained that this string was to be tied to the entrance of the labyrinth and unrolled as Theseus ventured deeper into it. As it turned out, this hack allowed our hero to navigate the maze, slay the Minotaur, and trace his steps back to freedom. Ariadne’s string is the key to this myth, so let’s unpack its role further.
Between Adaptability and Efficiency
On the face of it, the string is an absurd way of finding your way around in a labyrinth. Why not use a map? Or, absent a map, a series of “take the first right, walk straight 20 paces and turn left” instructions? All Theseus would have to do is trace his movements according to the plan and never stray from it. After all, Daedalus was the maze builder and presumably remembered its construction plan. One would expect a legendary techno-craftsman to produce some intricate contraption showing the way to Theseus. So, why the crude and simple string?
Simply put, because Theseus was about to be punched in the face.
Knowing that, Daedalus could not have given him a plan of the maze or a complicated contraption. What if Theseus loses the map, forgets the detailed instructions or the intricate contraption gets broken in his fight with the Minotaur? Ariadne’s string exemplifies the optimal balance between maximum adaptability and simple efficiency for a given system’s scale. The string doesn’t show the way forward or the maze’s layout. The string is dumb. Worse, it has nothing to do with the maze at all! It simply adapts to and interfaces with every twist and turn of the labyrinth while being highly efficient in showing Theseus only one simple thing – the path he took.
You see, adaptability is a function of a system’s ability to perform many possible simple actions more or less independently of each other. Ariadne’s string is an adaptability hack for the complexity of the labyrinth – it could interface with all possible permutations of that space. Multiple possible actions at the smallest of scales.
Efficiency, however, is a function of the ability of various system parts to work together to perform tasks at the largest possible scale. Ariadne’s string was expected to perform only one task at the scale of Theseus traversing an impossible maze, fighting a monster inside it, and getting out. The simplest of tasks at the largest of scales.
Designing a system for efficiency and adaptability is far trickier than it appears at first. Imagine a city’s transportation system designed for maximum efficiency: a network of trains and buses running on a tight schedule, minimal wait times, and optimized passenger carrying capacity. As is usually the case, such a system would be given as an example of optimal efficiency due to its just-in-time predictability and low operational costs. However, the system’s rigidity becomes apparent when a sudden, unexpected week of heavy downpours disrupts its operations.
While optimal under normal conditions, its efficiency doesn’t allow for quick adaptation to the new challenges posed by the heavy downpours. The transportation system, optimized for specific operational conditions, struggles to provide alternative routes or modes of transport that accommodate the change, leading to delays, congestion, and chaos. In other words, the system is optimized for a well-defined operational envelope, but that very optimization deprives it of the resources to quickly adapt to a dramatic change in the envelope.
Notice that if the system had a spare fleet of otherwise redundant minibuses, it could adjust to the sudden change in conditions much better. However, that same redundant fleet of minibuses – representing adaptability – will present extra costs and additional and unnecessary complexity in all ordinary conditions. There is a lesson here.
Alternatively, consider a modern tank battalion advancing in enemy territory with infantry support, acting according to doctrine as an efficient complex system. With more than a century of deployment history, tanks are shockingly efficient in interfacing with most questions the typical enemy can ask of them. Their interactions are usually short and have great finality in execution.
However, this particular enemy has deployed a swarm of dirt-cheap first-person view (FPV) drones, each armed with an armor-piercing warhead. As the swarm maneuvers at speeds exceeding 100 kilometers per hour, slamming into and destroying tank after tank, the advancing complex system has no way of interfacing with the drones. A punch to the face and a knockout. True story. What is the lesson here?
A system’s effectiveness is contingent on its ability to provide a distinct response to each environmental possibility it may encounter.
If a system cannot interface with the changing conditions in its environment, it will fail to be effective and, absent a transformation, will ultimately collapse. However, we have to remember that as long as the system keeps operating under the conditions for which it is optimized, it will be within its maximum effectiveness envelope. This is why so many highly fragile systems seem to operate just fine when viewed from the outside. This is also why the easiest way to derail a highly efficient system is to change the scale of its operational envelope. Even a slight shift in external conditions would often completely derail a highly optimized and efficient system.
The problem is that highly efficient systems lack the flexibility to adapt to new challenges. Like the transportation system discussed above, to become highly efficient, they need to remove all unnecessary complexity and redundancies, streamlining processes for optimal performance conditions. The very optimization that makes a system highly efficient prevents it from quickly adapting to change. Any highly efficient system is also highly fragile.
That is because, as I already mentioned, adaptability is a function of a system’s ability to perform multiple distinct actions at small scales. In other words, adaptability emerges when a system can interface with reality in multiple, often rare, non-optimal conditions. That fleet of minibuses is a small-scale redundancy, increasing the complexity and costs of the transportation system but allowing the system to adapt to the rare occurrence of a week-long downpour or other sudden disruptions. Similarly, installing radio frequency jammers on each armored vehicle is a small-scale redundancy, increasing the complexity and cost of a tank battalion but allowing the system to at least partially adapt to the sudden occurrence of an FPV drone swarm attack.
Highly adaptable systems, on the other hand, can interface with multiple environmental challenges but struggle with scaling up. As they grow, the costs of maintaining their complex adaptability increase to a threshold beyond which they cannot perform their actions efficiently at a given scale. In other words, past that scale threshold, the highly adaptable system has no other option but to optimize its processes for efficiency. Either that or the rising complexity costs at larger scales bring the whole system down.
Therefore, to be highly adaptable, a system has to either stay below a specific scale of operations or keep its adaptable elements small while growing in scale with a much more efficient structure and output. We will explore this dynamic further.
The time and scale trade-offs
If you think through this dynamic, you will notice that adaptability adds costly complexity here and now but may save a system in the future, while efficiency lowers costly complexity today but will doom a system in the future. This is why most systems will naturally drift towards increased efficiency at the cost of lowered adaptability. Optimizing for efficiency saves system resources here and now while optimizing for adaptability does not generate immediate effects for most systems.
That fleet of minibuses represents ongoing costs the transportation system must pay in the present, while the adaptability it provides may save the system in the future. Conversely, the efficiency of a transportation system streamlined for optimal conditions lowers its operational costs in the present but invariably renders the system more fragile to potential future shocks.
There is a time trade-off where the investment in adaptability – though burdensome with its added complexity and immediate costs – acts as insurance against future uncertainties. On the other hand, focusing on efficiency streamlines operations and reduces overhead in the short term while rendering the system brittle and unable to cope with unforeseen changes.
Again, adaptability increases your system’s costs in the present – think time, money, energy, and cognitive load – at all scales where it appears but allows your system to evolve at those scales. Efficiency saves you costs in the present – think time, money, energy, and cognitive load – but increases your system’s fragility at all scales.
In essence, adaptability buys resilience at the expense of current simplicity, while efficiency buys simplicity in the present at the cost of future vulnerability.
This is the efficiency versus adaptability trade-off represented along a time axis. However, the choice between adaptability and efficiency is also a problem of scale. There is an inherent trade-off between the number of ways a system can interface with reality and the scale at which it can coordinate these engagements.
The more complex a system’s actions, the higher the cost of performing them at a larger scale.
For example, consider the cost of deploying a hundred drone countermeasures locally in one sector of the front as opposed to hundreds of thousands across the structure of an entire army. The simpler the actions, the easier they are to perform at a large scale.
In other words, as I mentioned above, there is a scale threshold beyond which a system will be unable to perform complex actions without sacrificing the efficiency of its operations. While efficiency is about optimizing for a task at the maximum system scale, adaptability is about redundancies for rare tasks at multiple small scales.
The fundamental scale trade-off means that a complex system optimized for adaptability will have greater complexity at smaller scales, while a complex system optimized for efficiency will have lower complexity but operate at much larger scales.
We are witnessing these trade-offs today as FPV drone swarms obliterate thousands of tanks and armored vehicles on the fronts of Ukraine. FPV drones have been a known technology for more than twenty years and have been used in warfare for at least a decade, yet no modern military has fully adapted to them. So far, neither the theory of mechanized warfare nor command structures or individual tank designs can effectively interface with the complex questions asked by FPV drone swarms.
Modern militaries are systems optimized for efficiency in the present at vast scales – maneuver warfare, capturing territory, and access denial. Meanwhile, drone swarms pose a question of complexity at multiple small scales – exactly where modern militaries are highly efficient and cannot adapt quickly.
Again, as a system’s actions become more complex in adapting to potential environmental changes, its capacity to coordinate them effectively and scale up diminishes. A company might be very efficient in producing a limited number of complex widgets, but scaling up production would increase costs beyond the threshold at which it can make them efficiently. This trade-off presents a critical challenge for any system navigating between complex objectives and maintaining the ability to operate at a larger scale.
For example, consider the operational differences between special forces and conventional army units. Special forces perform highly complex tasks, such as covert surveillance, infiltration, sabotage, and engaging valuable enemy targets. These tasks involve constantly evolving and sophisticated technologies, highly specialized skills, precision, and adaptability. The complexity of special forces acts as a hard-coded limit on their size and scale of operations in maintaining coordination and effectiveness.
Conversely, conventional armies are designed and trained to engage in large-scale operations such as capturing territory and access denial. While these operations require coordination and discipline, they rely on the repetitive execution of much simpler tasks performed across large units, enabling them to achieve objectives on a grander scale.
The same dynamic can be observed in the contrast between startups and multinational corporations. Startups thrive on rapid innovation, fast feedback loops, iteration, and agility, focusing on developing new products or services. They are systems optimized for maximum adaptability at multiple small scales. Each of their elements is usually highly complex and adaptable to dynamically shifting operational envelopes.
For example, think of the ambiguity of startup position descriptions. Founders and their first employees must work across the entire operational envelope of the system – from sales to coding and management. Talk about extreme complexity at small scales. Their focus on adaptability and innovation requires flexibility and rapid decision-making that cannot be maintained as the organization grows.
As startups scale up into larger enterprises, the complexity of their operations invariably must decrease to standardize processes and achieve economies of scale. Startups buy the ability to operate at larger scales by reducing their complexity at those scales. Over time, the efficiency drift I mentioned above becomes inevitable for most former startups. They optimize for efficiency to unlock economies of scale while simultaneously curtailing adaptability.
This is why, while multinational corporations are present at global scales, they seriously struggle to innovate at the same speed and creativity as startups. It is much cheaper for a large corporation to buy startups than to maintain costly adaptability. The time and scale trade-offs to coordinating complex actions are unavoidable.
Ariadne’s string principle
Now, let’s return to Daedalus’ gift to Ariadne. The string has the optimal complexity required to interface with the labyrinth and the optimal efficiency to allow Theseus to retrace his steps. Daedalus’ genius lies in matching the complexity of the maze – that is, the questions it might ask of Theseus – and reframing the task of finding a way out as a simple problem of retracing steps. He addresses the time and scale trade-offs between adaptability and efficiency by focusing on performing the simplest and most efficient action in the present while matching the complexity scale of the maze. This is what Ariadne’s string principle is all about.
For maximum efficiency in its environment, a system must simplify its present actions while aligning with or exceeding its environment’s complexity at the corresponding scales.
Put differently, Ariadne’s string principle requires a system to perform two, at first sight, divergent maneuvers dynamically.
A system must streamline operations in the present while ensuring they match or exceed the complexity of the realities they may interface with.
Moreover, what represents optimal efficiency at a given scale of external complexity today will probably not work well tomorrow. Adaptability is future-oriented, and the principle demands that potential future operational envelopes be accounted for in efficiency calculations at all scales. After all, Ariadne’s gift to Theseus wouldn’t be any good for him if it didn’t account for a potential punch to the face.
Ariadne’s string principle dictates that a complex system must incorporate evolutionary adaptation across its elements by allowing continuous parallel small-scale experimentation at all scales where it interfaces with external conditions. In practice, this means that a complex system such as a corporation must allow its units interfacing with external conditions to undergo continuous evolutionary adaptations at their corresponding scales. Crucially, for this experimentation to benefit the whole system, successful adaptations achieved by these units must be communicated and replicated across the system.
Stability
However, there is a catch. Within simpler systems, with more or less streamlined operations, successful adaptations in one element can be replicated across the system without much instability. Think startups pivoting to a new direction. Their small size, the flat, networked structure of their organization, and the complexity of their units allow them to adjust to evolutionary adaptations quickly.
Not so with complex systems, which are usually structured hierarchically with a distinct center of control and coordination and elements optimized for efficiency. Hierarchies are very poor at dynamically augmenting their structure and operations in response to a change in external conditions. This is why evolutionary adaptability and experimentation at smaller scales increase instability within complex systems over time.
External conditions invariably change dynamically (think FPV drones), and to match the changing complexity of their operational envelope at various scales, a complex system’s internal structure has to evolve at a similar or faster speed. What good is an otherwise effective tank brigade if it cannot evolve and adapt at the same speed and complexity as the FPV drones attacking it? Remember – Ariadne’s string principle requires the system to match or exceed the complexity of its environment at the scales it interfaces with it. The problem lies in the adaptation sync.
Complex systems usually struggle to sync the adaptation pace of their constituent elements, causing instability over time.
This is why successful complex systems must ensure all their elements continuously engage with and adapt to external conditions at various scales while communicating and replicating successful adaptation strategies. The faster successful evolutionary adaptations can be transmitted and replicated across a complex system, the better it adapts to changing conditions and the more stable it is. This is why high command, senior management, and C-suite executives should always be fully involved at the same scales as their frontline units. Their participation only speeds up the percolation of successful adaptations across the complex system.
Consider this dynamic in practice. After more than two years of incessant FPV drone warfare, most frontline units on either side of the war in Ukraine have fully or partially adapted to the immediate danger of FPV drone attacks. These are usually small-scale, locally improvised adaptations of varying complexity – from shields welded to the tank to amateur radio jammers. However, these successful local adaptations have not been replicated by all elements of the complex systems of the opposing armies.
For such an adaptation to occur, the internal structures of both armies would have to evolve at the same pace and align with the complexity scales of conditions on the front. Senior commanders would have to reframe operational plans around the danger of FPV drones, and the successful adaptations of frontline units – from evolving structure to new technologies and tactics – would have to be replicated across all elements of each army. This poses a profound structural challenge to the current stability of these systems.
In fact, a certain kind of complex system, whether an army, a corporation, a university, or an authoritarian state, might prioritize an arbitrary internal stability state optimized for a given macro efficiency scale over adaptability at multiple smaller scales. In practice, these types of systems would consciously opt to avoid evolutionary adaptations at multiple small scales, choosing to maintain the arbitrary stability of their current structures and operational modes at a given macro scale. At what cost, you might ask? What a good question!
In a vacuum, such systems would quickly collapse from the internal build-up of entropy caused by the exponentially rising costs of non-adaptation to external conditions. In reality, such systems pay the rising costs of arbitrary stability by increasing resource consumption. As simple as that.
An army would throw more people into the grinder – the war in Ukraine is a grotesque illustration of that. An authoritarian state would expropriate as much as it can from its subjects. Many such cases! A corporation would eat up all investor cash and bank debt it can access while appearing as stable as a rock from the outside. The market favorite!
As long as a complex system can access additional energy sources, it can afford to opt for an arbitrary present stability state of optimized efficiency. This allows the system to optimize its operations for efficiency at a given macro scale while paying the costs of mal-adaptation at smaller scales. In other words, access to unlimited resources allows the system to ignore the future (adaptability), focusing on an arbitrary present it has optimized for (efficiency).
However, when additional energy sources dry up, such a system has to abandon stability and start its adventures searching for Ariadne’s string. It’s either that or a punch to the face and a knockout.
