• Construct expert benchmarks: Build and validate real-world underwriting models, return analyses, and investment cases used to evaluate frontier AI systems. • Stress-test model reasoning: Diagnose weaknesses in AI-generated real estate analyses, identifying where logic, assumptions, or market intuition fail. • Design frameworks: Translate how REPE professionals evaluate acquisitions, structure transactions, and manage assets across the investment lifecycle into problems that push the limits of AI reasoning. • Compare human vs. machine judgment: Systematically evaluate divergence between professional real estate judgment and AI outputs across asset classes, risk profiles, and fund structures.