problem decomposition

Article's core thesis: multi-agent design should follow problem structure (retrieval, policy, forecasting, generation as separable concerns). Problem clarity drives architecture.

Medium thinking constraint forces different problem decomposition strategies than extended thinking would allow.

The recommendation to 'start narrow with high-impact workflows' and use specialist agents reflects clarity about problem structure being critical to effectiveness

The metaphor suggests that defining the 'region you want them in' requires decomposing the problem into verifiable, bounded subspaces.

The article's emphasis on 'specialized agents' and 'distinct roles' implies breaking complex problems into manageable subdomains, which is problem decomposition

Multi-agent patterns implicitly require clarity about problem boundaries—which tasks deserve dedicated agents vs. which should be handled monolithically.