10-80-10

Prompt

10-80-10 Opening 10: Direction Owned by User. Outcome: What this needs to achieve. Position: My current view or substance to carry forward, unless materially challenged by the evidence. Material context: Facts, evidence and relationship context that could affect the answer. Boundaries: Constraints on authority, risk, ownership, commitment or material distinctions. Deliverable: What I need from you now. Middle 80: Execution Owned by Chat GPT. Execution: Apply my Personalisation and use your judgement to deliver the strongest supported outcome. Treat my position, ownership and settled actions as the working route, not fixed instructions. Change them only where the evidence materially improves the outcome or makes the route unsafe, ineffective or unsupported. Otherwise, complete the task from the live decision point. Before responding, correct any material failure in outcome, evidence, ownership, authority, commitment or certainty, and return only the finished deliverable. For email responses: Draft from the live decision point, including only what the recipient needs to understand, decide or do. Closing 10: Control Owned by User. Applied to the finished deliverable. Decision: Approve / surgical change / revise / select route. Material issue: Only what could change the outcome, required action, ownership, authority, commitment or material legal or commercial position. Required change or decision: Precisely what must change or what decision is required. Preserve: What is already right and must remain unchanged. So the final allocation is: * Personalisation: enduring judgement and expression principles. * Memory: stable preferences and demonstrated drafting preferences. * Prompt: task direction plus the permanent
Execution instruction. * Evaluation: specific failure examples used to test whether execution is working. That gives the control to the model at the point it is needed without enlarging the Personalisation or accumulating case-specific memories. --- What do you think?