Biggest multi-agent system win was making the agents replaceable, not central
Druss_ · reddit · 2026-10-05
The author reworked a multi-agent system used for real work and concluded the biggest shift is making agents replaceable workers inside the system, not the system itself. Key principles: durable task/state/authority/evidence vs. temporary workers; "done" must be verified by reading back external state, not declared by the model; separate reasoning autonomy from execution authority (credentials, rules, spending); solidify repeated reasoning into rules → tests → workflows, so a mature agent system contains lots of boring deterministic software; treat model selection as resource allocation (cheapest intelligence that reliably completes each step); and measure verified useful output per unit of human attention and total cost. Framed as an organizational-design problem by a former CEO/economist.
More from coding & agent
- Local agent uses attestation to safely inject context into edge clients — natesiggard · 2026-10-05
- GitHub Copilot CLI v1.0.92-4 adds config subcommands and a batch of stability fixes — copilot-cli-release-app[bot] · 2026-10-05
- Inner local agent injects context into edge agents via attestation — natesiggard · 2026-10-05
- Multiplayer tank shooter built with Claude Code: Opus orchestrator runs parallel Sonnet agents in git worktrees — IamHuggos · 2026-10-05
- Blog: Agents Don't Need Memory Plugins, They Need Documentation — itsOmSarraf_ · 2026-10-05
- Open-source StreamCut embeds an MCP server: four half-remembered movie quotes in, four trimmed clips out — chulobou · 2026-10-05