Mental Model

Overview

The desk is easiest to hold as five layers. AI sits in Research and Trade; Schedule, Brief, and Observe are timer, context load, and UI/infra. Anthropic names from Building effective agents (PDF).

Applied workflows

• Parallelization. Traders Warren, George, Ray, and Cathie run in the same tick via asyncio.gather — independent peers, no supervisor handoff.
• Orchestrator-Worker. Each trader owns the decision; a nested Researcher tool scrapes Brave/Fetch (and memory) and returns a brief — it does not place trades.
• Not used here (yet). Routing, Prompt Chaining, Evaluator-Optimizer.

Layers

• 1 Schedule. Floor checks is_market_open() (unless override), then wakes all four traders.
• 2 Brief. Each trader loads strategy and account resources, then builds a trade or rebalance message.
• 3 Research. Nested Researcher plans searches, hits Brave/Fetch, may write libsql memory, returns a brief.
• 4 Trade. Trader looks up prices (local market MCP or Polygon), then buy_shares / sell_shares or rebalance / change_strategy on accounts.
• 5 Observe. Push via Pushover; LogTracer writes traces; Gradio polls SQLite for board, P&L, and logs.

Layer map

Layer stack for one scheduled tick. Diagram key: ✦ AI agent · second line = Anthropic workflow · purple = actor · white = service.

---
config:
  theme: default
  themeVariables:
    background: "#ffffff"
    fontSize: "14px"
---
flowchart TB
    person[Person]
    schedule["1 Schedule
wake when open"]
    brief["2 Brief
strategy and account"]
    research["✦ 3 Research
Orchestrator-Worker"]
    trade["✦ 4 Trade
Parallelization"]
    observe["5 Observe
push logs dashboard"]

    schedule --> brief --> research --> trade --> observe
    person -->|watch board| observe

    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 research,trade agent
    class schedule,brief,observe service
    class person actor

Read layers top to bottom; drill into Traders for MCP belts and trade vs rebalance.