• Correct structure should make runtime behavior legible, not just prevent invalid states.
• Effects must remain tied to the state that created and owns them.
• Represent effects as commands plus receipts that record what changed and why each command exists.
• Use uncommitted transactions as dry runs that show consequences before applying them.
• Receipts explain the past; dry runs preview the future.
• Legible cause and effect lets people and agents diagnose or operate a system without guessing.