Andrew Ng’s 12-page playbook maps four design patterns for multi-agent systems
theomitsa · x · 2026-07-29
Andrew Ng's 12-page note lays out four stacked design patterns for agentic systems:
- Reflection: have the agent review its own work against a rubric for an immediate quality lift.
- Tool use: make the agent verify real data instead of guessing.
- Planning: let it reroute when something breaks instead of failing silently.
- Multi-agent collaboration: split roles so different checks catch different errors.
The accompanying diagram frames this as a graph-grounded multi-agent architecture built on a shared knowledge graph, positioned as a practical path from simple loops to production workflows.
More from coding & agent
- 'LLM code is sloppy' rebutted with real ffmpeg code: identical if/else branches in a top library — kingslyj · 2026-10-16
- Office-editing tool calls need file, worktree and unit: lessons from dsh-univer-office — FunDistribution6537 · 2026-10-16
- Architect MD workflow: Gemini one-shot outputs that beat Opus, via 5-lens plan critique — oikk01 · 2026-10-16
- Prove a coding agent actually ran the tests: check artifacts, not approvals — ComprehensiveShake76 · 2026-10-16
- Building an acceptance harness for AI spreadsheet edits: allowlists, recalc checks, idempotency — ukdicha-modakkk · 2026-10-16
- Agent actions need checks outside the model: NVIDIA demos lean on guardrails amid wave of real agent incidents — Creamy-And-Crowded · 2026-10-16