I built 174 AI agents that predict the future by fighting each other

Most multi-agent systems make agents cooperate. I made mine fight.

The Problem

Every prediction tool tells you what the crowd thinks. None of them tell you where the crowd is wrong.

The Solution: BlackSwanX

An adversarial intelligence engine where 200 citizen agents argue while a BlackSwan Assassin tries to kill the consensus. Runs 100% locally on Ollama. Zero API cost.

How it works:

1. Crawl — 5 free sources (DuckDuckGo, Reddit, HN, YouTube, Twitter)
2. Assassin's Mark — phi4:14b finds the Kill Shot before citizens start
3. Shadow Swarm — 200 citizens react with biased, emotional opinions
4. Cognitive Dissonance Matrix — calculates where belief diverges from reality
5. Decision-Ready Map — Linchpin + Antifragile Play

Example: "Will NVIDIA crash when the AI bubble pops?"

The system activated 20 agents (Economist, Quant Analyst, Panic Seller, Chaos Mathematician...) and found:

• Kill Shot: Quantum computing making GPUs obsolete (10% probability)
• Citizens: 25% bull / 65% bear
• Dissonance: 33.6/100 — MAXIMUM CHAOS
• Antifragile Play: Diversify into quantum computing partnerships

The 3-Model Strategy (runs on 16GB MacBook)

Role	Model	Purpose
Swarm	llama3.2:3b	200 biased citizens
Assassin	phi4:14b	Kill shot reasoning
Nexus	mistral-small:24b	Synthesis + DAG

Self-Learning (SONA)

After every run, SONA audits all agents:

• Boosts citizens that caught risks others missed (2x weight)
• Demotes ones that missed critical threats (0.3x)
• Stores patterns in a ReasoningBank
• The more you use it, the smarter it gets

174 Expert Agents

Including a Chaos Mathematician, a Vedic Astrologer, a Panic Seller, a Street Smart Hustler ("your pitch deck is pretty, show me your bank account"), and a Gen Z Culture Decoder.

Quick Start (2 minutes)

bash
git clone https://github.com/Kalki-M/BlackSwanX.git
cd BlackSwanX
ollama pull llama3.2:3b && ollama pull phi4:14b
pip install -r requirements.txt
bash start.sh