Agent performance depends on harness design, not just the model

Original: 📝 Learning note: What is an agent harness?

Source: x.com ↗