Keeping context and decisions consistent across parallel AI agents

Short summary

Running multiple AI coding agents in parallel creates a coherence problem: each agent's context is isolated, so decisions, conventions, and interface changes made by one agent never reach the others. Three types of state drift — durable decisions, current contracts, and in-flight file ownership — each require different mitigation strategies. The highest-leverage fix is settling shared design and interfaces before fan-out, then giving every agent a common, writable home for durable decisions.

• •Parallel agents drift because per-session context is never shared across worktrees
• •Three drift sources: durable decisions, current interface contracts, and in-flight file ownership
• •Best practice: resolve interfaces upfront and maintain a shared decision file all agents read and write

Generated with AI, which can make mistakes.