How a swarm of 10，000 agents solved Navier-Stokes
OpenAI researcher Noam Brown explains how a minimally scaffolded swarm of 10,000 agents used 130 billion tokens over 88 hours to crack Navier–Stokes, while mapping multi-agent test-time scaling, RSI bottlenecks, and alignment risks from cooperative scheming and fragile chain-of-thought monitoring.