Show HN: Lessons learned from running Claude Code swarms at scale
A developer has shared lessons learned from running large-scale coding agent swarms using a tool called 'fleet'. The primary challenges encountered involved inefficient token usage and context management with Claude's existing abstractions like CLAUDE.md and skills. The developer suggests a hierarchical knowledge base as a more scalable alternative and highlights the need for custom tools to enable user interaction in background sessions. AI
IMPACT Provides practical insights for optimizing the use of large language models in agentic workflows.