Two research-grade benchmarking platforms launched: CivBench (multi-agent game environments with live leaderboards) and A real (LLM Skirmish, a time-strategy game for agent evaluation). Both are open-research with potential future B2B monetization via enterprise model evaluation licensing. Gap: benchmarks remain siloed by use case; no unified protocol for comparing agent performance across reasoning, code generation, and real-world task execution.