Multi-agent AI systems fail at coordination, not intelligence

Original: a databricks tech lead just spent 26 minutes on the part of multi-agent nobody wants to say out loud:

Source: x.com ↗