March 2026

Swarm

A multi-agent market sentiment simulator. Feed it a news headline and a swarm of AI trader personas — quant funds, macro tourists, retail degenerates — react independently, then debate each other and update their positions.

The Problem

Market-moving events are ambiguous. A Fed rate hike is bullish for banks, bearish for growth stocks, and neutral for commodities — depending on who you ask. No single analyst captures all of that. The interesting signal is not any one view; it is where a diverse set of participants land after they have heard each other out.

How It Works

Swarm runs 15 distinct trader personas in parallel — each with a different role, style, natural bias, and time horizon. A quant fund does not react the same way as a macro tourist or a retail momentum trader.

Round 1 is independent: every agent reads the headline and forms a view without knowing what anyone else thinks. Round 2 is the debate: each agent reads a summary of all other positions and decides whether to hold firm, partially update, or flip entirely. View changes are tracked and surfaced in the final output.

Why Two Rounds

A single round of parallel calls gives you breadth of views but no interaction. The interesting dynamics in real markets happen when participants update on each other. A momentum trader who reads that every macro fund is bearish might temper their enthusiasm. A contrarian who sees consensus forming in one direction has more reason to go the other way.

Round 2 is what makes the output feel alive rather than just a list of opinions.

Design

The entire tool is a single Go binary with no runtime dependencies. Agents run concurrently via goroutines. The output is a color-coded terminal display with a signal bar chart, sentiment breakdown, and a watch list of the most-mentioned assets.

Install via the releases page or go install — no config files, no accounts, just an API key.

Working on something in this space?

Get in touch →