Worker Agents

Overview

Layer 4 Ideate on the Mental Model. Worker Agents are RoutedAgent instances (often Creator-generated) that propose Agentic AI business ideas. On each message they may bounce the draft to another agent before returning; world.py writes ideaN.md.

Behavior

• Persona system message. Interests, optimism, and risk appetite shape the idea style (template defaults to Healthcare / Education disruption).
• Probabilistic peer bounce. About half the time (CHANCES_THAT_I_BOUNCE_IDEA_OFF_ANOTHER = 0.5) the worker forwards the idea via messages.find_recipient() for refinement.
• Prompt Chaining. Draft → optional peer refine → return content to Creator / world. See Building effective agents.
• Batch spawn. world.py creates many agents in parallel; each successful reply becomes an ideaN.md file.

Worker flow

One worker receives a create / idea message. Diagram key: ✦ AI agent · second line = Anthropic workflow · purple = actor · white = service.

---
config:
  theme: default
  themeVariables:
    background: "#ffffff"
    fontSize: "14px"
---
flowchart TB
    world[world or Creator message]
    worker["✦ Worker Agent
Prompt Chaining"]
    draft[Draft idea]
    peer["✦ Peer worker"]
    out[idea markdown]

    world --> worker --> draft
    draft -->|bounce chance| peer --> out
    draft -->|no bounce| out

    classDef agent fill:#e8f0ee,stroke:#0f3d36,stroke-width:2px,color:#0d0d0d
    classDef service fill:#fafaf8,stroke:#c8c8c2,stroke-width:1px,color:#0d0d0d
    classDef actor fill:#f3eef8,stroke:#6b4f9a,stroke-width:2px,color:#0d0d0d
    class worker,peer agent
    class world,draft,out service

Template and runtime behavior: 5_autogen/agent.py; spawn loop: world.py.