Entropy
Entropy: Systems naturally drift toward disorder without continuous energy input to maintain structure. Code gets messy without refactoring; organizations drift without active culture maintenance; relationships deteriorate without investment. The natural direction is always toward disorder β order requires effort.
What Is Entropy?β
In thermodynamics, entropy is a measure of disorder in a closed system. The Second Law of Thermodynamics states that in any isolated system, entropy increases over time β order naturally degrades into disorder. Heat flows from hot to cold; ice melts; organized structures break down. Creating or maintaining order requires energy input from outside.
The concept generalizes to any complex system that requires active maintenance to remain organized:
Software entropy (software rot): Code that works correctly today degrades over time without maintenance. Dependencies become outdated, security vulnerabilities accumulate, the surrounding system changes but the code doesn't, and technical debt compounds. Even without a single person making a bad change, a codebase becomes progressively harder to maintain through inaction.
Organizational entropy: Processes, cultures, and standards that were intentionally designed drift over time. The hiring process that was carefully calibrated three years ago has accumulated informal exceptions. The product vision that was crisp at the founding has been modified by a hundred small decisions. Documentation is stale. Team norms have shifted.
Physical entropy: A clean room becomes dirty without cleaning. A garden becomes overgrown without tending. A well-maintained machine degrades without servicing.
The practical lesson: maintenance is not overhead; it is the cost of maintaining the value of something you've already built. Systems that are "let run" without active maintenance are degrading β the question is only how fast.
Three Real-World Examplesβ
Technical Debt Accumulationβ
A SaaS company ships a product with clean, well-documented code. Over two years of rapid feature development, the team takes shortcuts (technical debt) β messy abstractions, duplicated code, untested components β in exchange for speed. Without regular refactoring cycles, entropy compounds: the messy areas attract more mess (it's harder to write clean code in a messy context); documentation falls further behind reality; test coverage shrinks as a percentage of the codebase.
After three years, velocity has fallen dramatically not because the team is less capable but because the software has entropied. The cost of adding a feature is 5x what it was at year one, because every change requires understanding and working around accumulated disorder.
The fix is regular investment in reducing entropy: refactoring sprints, documentation updates, test coverage requirements, dependency updates. These feel like overhead because they don't produce new features β but they are what maintains the value of the existing investment.
Organizational Culture Entropyβ
A startup founding team creates a strong culture: direct feedback, intellectual honesty, fast decision-making, customer obsession. For the first 50 people, the founders can model and reinforce this culture directly. At 200 people, new managers who were hired and promoted during rapid growth are modeling a diluted version of the culture they received during onboarding. At 500 people, the culture has drifted significantly from its origin.
This is entropy: without active investment in transmitting, reinforcing, and occasionally restoring the culture (through training, firing for culture violations, celebrating culture exemplars), it degrades. Culture entropy is the main mechanism by which companies "lose their culture" as they scale.
The Maintenance of Physical Infrastructureβ
Roads, buildings, and bridges require continuous maintenance to remain functional. A road not resurfaced deteriorates from cracks to potholes to structural failure. The cost of repair after each stage of deterioration is substantially higher than the cost of maintenance at the previous stage. The US infrastructure deficit β accumulated deferred maintenance β is an entropy story: decades of underinvestment in maintenance allowed infrastructure to deteriorate, and the catch-up cost is far higher than the maintenance cost would have been.
When to Use Itβ
β Use Entropy thinking when:
- Planning maintenance budgets for software, infrastructure, and organizations
- Explaining why "doing nothing" is not a neutral choice
- Diagnosing why quality, culture, or performance has declined despite no obvious cause
- Setting expectations about the cost of not-maintaining something
| Pairs well with | Why |
|---|---|
| Homeostasis | Homeostasis is active resistance to entropy β the cost of maintaining it |
| Resilience Thinking | Resilient systems invest in entropy management as part of design |
| Diminishing Returns | Maintenance investment has its own diminishing returns curve |
Three Real-World Examplesβ
Software Codebase Degradationβ
Every software project accumulates entropy. Quick fixes create technical debt; workarounds layer over earlier workarounds; documentation drifts from reality; dependencies become outdated. Without deliberate effort β refactoring, code reviews, architectural reviews β a codebase becomes progressively harder to change. Engineers spend more time understanding the existing mess than writing new features. This is entropy in action: disorder increasing without active maintenance energy being continuously applied.
Organisational Communication Breakdownβ
A 10-person startup has roughly 45 possible communication pairs. A 100-person company has 4,950. Without deliberate structure (processes, documentation, meeting cadences), communication entropy increases faster than headcount. Information siloes form; decisions get made on outdated information; different teams build incompatible systems. The disorder state for a large organisation is dysfunction; maintaining order requires continuous, expensive coordination effort.
Physical Infrastructure Decayβ
The US infrastructure backlog β bridges, roads, water systems β illustrates entropy at national scale. Without maintenance, physical systems degrade toward disorder: steel corrodes, concrete cracks, pipes rust. The energy required to maintain a system at current state is non-zero; neglect is not neutral, it is an active choice to allow entropy to operate. The longer maintenance is deferred, the greater the eventual restoration cost β often nonlinearly so.
When to Use Itβ
β Apply entropy thinking when:
- Building maintenance strategies for any system (software, physical, organisational)
- Evaluating why quality degrades over time in the absence of active management
- Designing for resilience β systems that resist entropy require energy inputs
β Be cautious:
- Entropy is a physical law; analogies to social systems are illuminating but imperfect
- Not all degradation is entropic β some is strategic (planned obsolescence) or competitive
- Entropy doesn't mean inevitable collapse β steady energy inputs can maintain order indefinitely
| Pairs well with | Why |
|---|---|
| Homeostasis | Homeostasis is a system's mechanism for resisting entropy |
| Resilience Thinking | Resilience is the capacity to maintain order under entropic pressure |
| Theory of Constraints | Entropy tends to concentrate in bottlenecks |
Common Misuses and Limitationsβ
Treating entropy as only physical. The thermodynamic law applies to closed physical systems. The analogy extends loosely to organisations and information systems, but the mechanisms differ. Don't over-engineer the analogy.
Using it to justify inaction. "Everything tends to disorder, so why maintain it?" Entropy doesn't justify neglect; it prescribes constant maintenance investment as the cost of keeping systems functional.
Ignoring open systems. Thermodynamic entropy applies to closed systems. Life persists precisely because organisms are open systems β they import energy and export disorder (heat). Organisations, too, can counteract entropy by importing resources and talent and removing dysfunction.
Related Modelsβ
| Model | Relationship |
|---|---|
| Homeostasis | Homeostatic systems actively resist entropy |
| Diminishing Returns | Systems under entropic pressure often show diminishing maintenance returns |
| Feedback Loops | Balancing feedback loops are entropy-resisting mechanisms |
Frequently Asked Questionsβ
Is entropy always bad?
In thermodynamics, entropy is neither good nor bad β it simply is. For complex systems like organisations and codebases, entropy (increasing disorder) is typically costly because ordered systems are more productive. But selective entropy can be valuable: destroying outdated structures (reorganisations, refactors) to make room for new, higher-order structures. Creative destruction is intentional entropy in service of rebuilding.
How much maintenance is "enough" to counteract entropy?
Enough to keep the system's disorder from increasing. For software, engineering teams typically aim for 20β30% of capacity allocated to maintenance and technical debt reduction (the "refactoring budget"). For organisations, it means regular process reviews, documentation updates, and talent renewal. The exact amount depends on the system's complexity and rate of external change β faster-changing environments require more maintenance to stay current.
What's the connection between entropy and innovation?
High-entropy (disordered) systems are hard to change beneficially because you can't predict the effects of interventions. Low-entropy (well-maintained) systems are easier to improve because their structure is legible. This is why technical debt slows feature development, and why organisational confusion slows strategic execution. Maintaining order is a prerequisite for effective innovation, not an alternative to it.
Further Readingβ
- Prigogine, I. & Stengers, I. (1984). Order Out of Chaos β thermodynamics and self-organisation
- Fowler, M. (1999). Refactoring: Improving the Design of Existing Code β software entropy and how to fight it
- Brooks, F. (1975). The Mythical Man-Month β entropy in software projects and organisations
Apply with AIβ
π Identify entropy in your systems and plan maintenance with MindMax β
This page is part of the MindMax Mental Models Knowledge Base.