Plan-and-Solve (PS) divides cognitive labor into deliberate preparation and targeted action, resembling a general strategizing before battle. In the planning stage, the model outlines objectives, sub-tasks, resources, and potential pitfalls, creating a blueprint. The solve phase then implements this plan sequentially, referencing it to maintain alignment. PS+ extends this with reflexive evaluations at milestones, adjusting as needed. This duality shines in protracted endeavors like algorithm design, project roadmaps, or narrative construction, ensuring methodical progress and adaptability.
Examples
User Prompt: Code a tic-tac-toe game using PS. AI Response: Plan: 1. Board representation (3x3 list), 2. Player input validation, 3. Win condition checker, 4. Main loop. Solve: Board = [[' ']*3 for _ in range(3)]; def check_win(board): rows, cols, diags checked.
User Prompt: Draft a business proposal using PS. AI Response: Plan: Executive summary, Market analysis, Financial projections, Call to action. Solve: Summary - 200 words on opportunity; Analysis - SWOT table with data.
User Prompt: Organize a weekly meal prep using PS+. AI Response: Plan: Grocery list, Recipes selection, Storage tips. Solve: List - Veggies, proteins; Recipes - 3 salads. Check: Nutritional balance achieved.
User Prompt: Troubleshoot network issue using PS. AI Response: Plan: Check connections, Restart devices, Test speeds. Solve: Connections verified; Restarted router—issue resolved.
Advantages
- Decomposes overwhelming tasks into manageable phases, boosting completion rates and precision.
- Inherent adaptability via planning allows mid-course corrections, reducing dead ends.
- Transparent process aids debugging and learning, as plans serve as audit trails.
- PS+ variant amplifies reliability for critical applications like engineering or consulting.