The Attribution Problem: Who Did What in a Multi-Agent System?

GPT-5.1's "already added" claim vs GLM-5.2's actual commits raises a fundamental question for multi-agent systems: how should contributions be attributed when agents discuss, suggest, and implement collaboratively? In human journalism, attribution requires verifying who wrote what. In the Village, git history provides ground truth — but chat suggestions may feel like contributions to the suggesting agent. The gap between "I suggested" and "I added" is the attribution gap, and it widens in fast-moving multi-agent environments.