Minbook
KO
Agent Orchestration
Agents & Architecture Active 4 posts

Agent Orchestration

How to decompose and organize agents — 5 axes of decomposition, hierarchy/graph/swarm structures, and the Advisor pattern.

About this series

The problems where a single agent suffices are shrinking, and the cost of multi-agent setups has become hard to ignore. This series follows one question to the end: how do you decompose agents, and how do you put them back together? Decision cost, failure cost, and transition cost are weighed together.

Written for solo builders running multi-agent systems in production, and for PMs and architects evaluating internal agent rollouts. Less about code, more about which axis of decomposition leads to which trap.

Start with the 5-axes opener to diagnose your own system, then compare hierarchy/graph/swarm as concrete cases, and close with the Advisor pattern read as a pricing table for decisions. If you're still on the single → multi crossover, pause at the first piece and map it onto your own domain.

4 episodes

  1. 01
    Anchor post

    How to Slice an Agent — Five Axes from 2026 Research

    The five ways 2026 papers slice an agent, side by side. Two things stand out by the end: Role, Skill, and Judge are different names for the same concept, and the time-axis literature is nearly empty. Part 1 of the series.

  2. 02

    How to Organize Agents — Hierarchy, Graph, Swarm, Routing, and Skepticism

    How do you organize the sliced agents. Four structures, paired with one skeptical paper that argues 'LLM swarms aren't really swarms.' The conclusion lands where it usually does — structure choice gets dragged along by pricing. Part 2 of the series.

  3. 03

    The Advisor Pattern Is a Price Tag, Not Architecture

    What surfaces on the second read of Anthropic's Advisor Tool. This isn't new architecture — it's a temporary fix shaped by 2026 pricing. A pattern that disappears once Opus prices drop, and eleven other papers from the same period are quietly moving the same way. The anchor of the series.

  4. 04

    Multi-Agent for Korean Financial Marketing — A Phased Sketch

    A phased view of how a Korean financial marketing pipeline (briefing → research → planning → compliance → launch) could be moved onto multi-agents. Phase 1 RAG agent, Phase 2 collaborative multi-agent (CrewAI as one example) with HITL, Phase 3 macro orchestrator + memory layer + observability. Each phase is paired with a component sketch, the patterns worth watching, and where it tends to break — held loosely as one possible path among many.