Understanding AI analyzes why agent clusters may become the next scaling law

Understanding AI wrote analyzing whether multi-agent collaboration constitutes a new scaling law. The article notes OpenAI achieved collaboration of thousands of agents by training models in multi-agent environments and using messaging tools for communication, such as 10000 agents solving the Navier-Stokes math problem in a few days; but Noam Brown admitted the multi-agent method contributed less than 10% to the breakthrough, with the main credit going to the underlying model.