New benchmark tests AI agents that build other AI agents

Original: Hyper-𝜏-bench: Evaluating agents that build agents

Source: sierra.ai ↗