Autonomous mode risks over-acting while copilot mode risks displacing the human performer

The two operating modes share the same ethos and diff-size ladder but are optimized for opposite failure modes. Autonomous mode must self-restrain to avoid acting too much (the runaway-agent risk); its decision table and wait-bias exist for this reason. Copilot mode must stay subordinate so the human’s liveness is never displaced; its propose-explain-wait loop and one-proposal-at-a-time rule exist for this reason. Both failure modes undermine the performance for the same underlying reason: the audience trusts that a human is in control, and either runaway automation or an agent dominating a human’s choices breaks that contract.

Examples

A copilot agent that queues three pending proposals and applies them in sequence has violated copilot subordination. An autonomous agent that saves two ideas in one cycle has over-acted and broken the one-concept-id-per-save rule.

Assessment

Describe the distinct risk each mode guards against and explain why both risks threaten the same underlying performance value.