Zero-Capital Economic Agent Swarm
Zero-Capital looks for income ideas that need no starting money. Scanner, Market, and Evaluator react to the same prompt at once. Writer waits until they finish, then turns the shared board into one plan.
What it does
A concurrent swarm of 5 AI agents (Scanner, Market, Evaluator, Writer, Observer) that lives in ONE Mozaik RuntimeState, receives the same message.sent event, and processes it simultaneously — independently inferring, calling real tools, and coordinating through a shared board. The goal: find the first real-money opportunity at $0 capital cost. Concurrency isn't a feature we claim, it's the core mechanic, proven with per-agent timestamps.
How the agents run concurrently
An operator sends one economic question -> broadcast as a single message.sent event. Scanner finds zero-capital opportunities, Market calls a real tool (get_quotes) for live BTC/S&P500 quotes, Evaluator ranks each by Expected Value (EV = P x reward - $0). All three write to a SharedBoard. Writer reads the full board and synthesizes ONE action plan. Observer logs every agent's answer with a timestamp — live proof of true concurrent execution. We use the actual @mozaik-ai/core v4 runtime: defineRuntime, createAgent, createHuman, SemanticEvent, runLoop, situation handlers, FunctionCallOutputItem tool-calling. Event-driven concurrency + adaptive coordination through shared RuntimeState — not a sequential pipeline in disguise.
