chapter eight

8 Collaboration: When one agent is not enough

 

This chapter covers

  • Deciding whether another agent will add useful capacity or only coordination cost
  • Passing work between specialists with Handoff Chain
  • Dividing independent work and reconnecting the results with Fan-Out/Gather
  • Testing a contested claim with Adversarial Review
  • Giving a manager responsibility for a multi-stage goal with Hierarchical Delegation
  • Measuring the seams where a team loses context, duplicates work, or hides disagreement
“The power of intelligence stems from our vast diversity, not from any single, perfect principle.”

— Marvin Minsky, The Society of Mind, p. 308

Anthropic tested a research agent on a task that sounds simple until you try to complete it: identify every board member of every Information Technology company in the S&P 500. One agent searched company after company. It could find correct names, yet it did not cover the full set. The task was too broad for one sequential search.

A lead agent changed the shape of the work. It divided the companies among subagents, let them search in parallel, and gathered their findings. Each worker had a bounded set of companies. The lead agent could see which sets had returned, which citations supported each name, and where coverage was still missing. The extra agents earned their place because the task contained genuinely independent pieces of work [1].

8.1 What collaboration changes

8.2 Pattern: Handoff Chain

8.2.1 Walking through an airline-support handoff

8.2.2 A chain in a coding agent

8.2.3 When Handoff Chain breaks

8.3 Pattern: Fan-Out/Gather

8.3.1 When apparent parallelism collapses

8.3.2 Gathering without hiding the minority

8.3.3 Parallel work in coding agents

8.3.4 When Fan-Out/Gather breaks

8.4 Pattern: Adversarial Review

8.4.1 Three mechanisms that look like review

8.4.2 Evidence from code review systems

8.4.3 Running a bounded review

8.4.4 When Adversarial Review breaks

8.5 Pattern: Hierarchical Delegation

8.5.1 What the manager and workers need from each other

8.5.2 Hierarchical delegation in coding agents

8.5.3 When Hierarchical Delegation breaks

8.6 Testing and observing collaboration

8.7 Composing collaboration patterns

8.8 The implementation layers: Where control lives

8.8.1 Three communication channels

8.8.2 MCP and A2A are the plumbing

8.9 Argus checkpoint: Collaboration

8.10 Summary