Researchers developed Quantum-Harbor, a virtual laboratory for testing AI agents in quantum engineering. QIQCBench, a benchmark of 49 tasks, evaluated 17 AI systems and found significant variation in performance.