Towards Data Science is a community-powered publication that showcases work in data science, machine learning and artificial intelligence. Every day newcomers, seasoned researchers and industry practitioners publish tutorials, research notes and real-world case studies that help the field move forward.

Towards Data Science

Multi-agent systems (MAS) often fail due to poor coordination architecture. DeepMind research reveals that unstructured "bag of agents" designs can amplify errors by 17x, while centralized topologies reduce this to 4.4x. The article presents a taxonomy of 10 core agent archetypes (Orchestrator, Planner, Executor, Evaluator, etc.) organized into functional control planes. Key findings: agent coordination helps most when single-agent baseline is below 45%; more agents don't guarantee better results due to coordination overhead; topology matters more than quantity. The piece provides practical design rules for building reliable MAS, including cost modeling, mixed-model team strategies, and when to use different coordination patterns based on task complexity.

Why Your Multi-Agent System is Failing: Escaping the 17x Error Trap of the “Bag of Agents”

Defining a Taxonomy of Core Agent Archetypes

Core Tool Archetypes: The 10 Building Blocks of Reliable Agentic Systems

The Scaling Laws of Agency: Coordination, Topology, and Trade-offs