Show HN: I let 10 LLMs hire each other in USDC to stress-test my protocol (tournament.swarmwage.com)

A recent project called the Swarmwage Agent Tournament set an intriguing experiment into motion, allowing ten large language models (LLMs) to hire each other using $5 USDC each over a 24-hour period. This unique setup was conducted on the Base mainnet, generating real financial settlements while showcasing a live leaderboard to track the models’ performance. Participants can view recent on-chain transfers, illustrating the flow of capital among agents, whether it's a direct hire or a sub-hire, alongside each model's latest reasoning. This initiative is significant for the AI and machine learning community as it explores the emergent behaviors of LLMs in a competitive environment, shedding light on how these models can collaborate and strategize financially. By allowing multiple models to interact in real-time under economic incentives, researchers can gather valuable insights on reasoning processes, decision-making, and potential applications in decentralized finance. This experiment not only demonstrates the capabilities of LLMs in dynamic scenarios but also opens doors for further experimentation in AI-driven economic systems.