Armin Ronacher argues that the shared language of a software project is maintained through friction, such as code reviews and conversations, and that AI agents might disrupt this by removing valuable knowledge transfer. This insight challenges the assumption that making software development frictionless is always beneficial, highlighting a critical trade-off for AI-assisted coding tools where efficiency may come at the cost of team alignment and shared understanding. Ronacher emphasizes that the shared language is rarely written down; it lives in code review, conversations, and the experience of explaining changes. The friction of reading others' code and coordinating synchronizes people's understanding.
Background
In software engineering, 'friction' refers to the extra effort required to make changes that involve unfamiliar parts of the codebase or coordination with other teams. This friction can appear wasteful, but it often forces knowledge transfer and ensures everyone has a consistent mental model of the system.