liveBook · Manning

Swarm

As explained in Vibe Engineering

A coordinated group of AI agents that divide or jointly execute parts of a larger workflow.

A swarm coordinates multiple agents on a larger workflow, but adding agents is not automatically an upgrade: a five-agent swarm beats one agent only for a narrow class of parallelizable work. section 5.0

Coordination also multiplies operational risk. Treat isolation as an architectural requirement: the intended design gives each agent a disposable environment and access only to the resources it needs. That remains a direction to work toward, not a solved guarantee. section 5.6.6

Use swarms selectively, after building the supervision and observability needed to manage agents. Coordinating several agents can mean coordinating several independently destructive processes. section 5.0 section 5.7

How Vibe Engineering covers it

How deep it goes
Limits the advantage to narrow parallelizable tasks and frames per-agent isolation as a future architectural need.
Examples
Contrasts one agent with a five-agent swarm and proposes isolated disposable environments for each agent.
What it assumes
Assumes accumulated operational maturity, including prior agent supervision and observability before orchestration.
What you can do after
Evaluate whether work is parallelizable and whether agents can receive scoped, isolated access.
How it connects
Connects agent swarms to agent orchestration, CI pipelines, and sandboxing.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Vibe Engineering. The section links point to the parts of the book it draws on.