Claude Code ran 5 agents at once in biggest upgrade

Anthropic tested Claude Code with five simultaneous agents on a complex task. The multi-agent setup completed the work faster than a single agent but introduced coordination overhead and occasional conflicts. The experiment highlights both the potential and current limits of scaling agentic AI workflows.