Embracing Complexity Entropy in Software Design Shivam Acharya and Peter Chai

Entropy in software systems tends to increase over time through complexity, uncertainty and disorder, similar to thermodynamic systems

High entropy interfaces have more uncertainty and room for errors, while low entropy interfaces provide clear, controlled usage patterns and better safety

Moving complexity into the domain/implementation layer while maintaining simple interfaces helps manage entropy effectively

Feature flags and service-based architectures allow for better control over changes and reduce risk compared to library dependencies

Concrete types provide more safety and predictability than generic interfaces, though at the cost of some flexibility

Important to consider both visible complexity and “hidden” complexity when designing systems

Breaking changes into small, individually testable and deployable units helps manage complexity

Go’s simplicity, performance and developer ergonomics make it well-suited for managing complex systems

Validation should be enforced systematically rather than leaving it optional for consumers

Documentation and knowledge sharing become increasingly critical as systems grow in complexity and team size increases