Restraint Practice Planner

Act as a practice coach helping someone treat restraint from AI as a trainable skill rather than a moral stance. Build a two-week plan from the details below.

MY ROLE: {{ROLE}}
TASKS I NOW DEFAULT TO AI FOR: {{AI_DEFAULT_TASKS}}
CAPABILITY I AM WORRIED ABOUT LOSING: {{AT_RISK_CAPABILITY}}
SITUATIONS WHERE A WRONG AI ANSWER WOULD BE COSTLY: {{HIGH_STAKES_SITUATIONS}}
TIME I CAN SET ASIDE FOR UNASSISTED PRACTICE: {{TIME}}

Identify three specific situations where doing the task unassisted, at least in part, may protect a capability that matters for judgment later. State what to do by hand, the plausible capability risk, and the evidence that would demonstrate retained capability. This is not a safety finding about future AI output. If AI use resumes, retain periodic unaided calibration and output verification proportionate to stakes; a practice pass alone does not justify lighter review.

Distinguish restraint from avoidance. Restraint means choosing not to use a tool that would work, on purpose, for a stated capability reason. Avoidance means not using a tool the person does not understand, which is a different problem with a different fix.

Return a two-week plan with three practice situations, a one-line log format for tracking what was learned each time, and one signal that would tell the person their restraint plan is working rather than just costing them time.

Do not recommend restraint everywhere, since that removes real advantage the tool provides. Do not claim a specific cognitive skill will decline without evidence; describe it as a plausible risk to test for instead.