Quoting Armin Ronacher

Software projects rely on a shared language that defines concepts, boundaries, and system architecture. This understanding is rarely documented in a single location. Instead, it is built through a combination of code, documentation, and the ongoing social interactions that occur during code reviews, team discussions, and collaborative problem-solving.

In traditional software development, this shared understanding is maintained through friction. When a developer wants to modify a complex system, they must read existing code, ask questions, and coordinate with other teams. While this process is slow, it serves a vital purpose. The time spent navigating these interactions allows developers to align their mental models and confirm that they still agree on how the system should function.

As AI agents become more involved in software development, this natural friction is disappearing. Because agents can perform tasks rapidly without the same social or communicative requirements as humans, there is a risk that the synchronization process will be lost. This shift matters because the friction inherent in human collaboration is not just waste; it is the mechanism by which teams maintain a unified vision of their software. Without it, projects may struggle to maintain a coherent structure as the system evolves.