DEV Community

showjihyun
showjihyun

Posted on

Prophet: We Built a Wind Tunnel for Marketing Campaigns

#opensource #ai #marketing #simulation



Focus groups lie. A/B tests are too late. Brand-lift studies take 6 weeks, cost $50K, and tell you nothing about why a message failed.

So we built Prophet — an open-source platform that simulates how your message spreads through 10,000 AI agents organized into the communities you actually care about.

💡 Think of it as a wind tunnel, but for campaigns instead of airplanes.

🔄 The Core Loop

⚙️
1. Set up
Message · communities · dials 		▶️
2. Run
10K agents · real-time 3D graph 		⚡
3. Intervene
Pause · inject crisis · replay
📊
4. Analyze
Adoption curves · sentiment 		🔍
5. Drill in
Read actual conversation threads 		⚖️
6. Compare
Clone · change variable · diff

🧠 What's Under the Hood

Component Detail
Agent Architecture 6-layer cognitive stack: perception → memory → emotion → cognition → decision → influence
Network Model Hybrid Watts-Strogatz + Barabási-Albert with validated clustering & modularity
Opinion Dynamics Deffuant bounded-confidence model for endogenous polarization
Emergent Detectors Viral cascade · Slow adoption · Polarization · Collapse · Echo chamber
Cost Control 3-tier LLM inference — full 10K-agent run under $5
Uncertainty Monte Carlo sweep with P5 / P50 / P95 quantification

🚀 Get Started 

git clone https://github.com/showjihyun/Prophet
cd Prophet
npm install
# No API keys required to start
npm run dev
Enter fullscreen mode Exit fullscreen mode

✅ Reproducibility Guarantee

Every claim in the README has per-step JSON you can audit.

Every equation traces back to a published paper.

Everything is reproducible. Runs on a laptop.

🤝 We're Looking for Contributors

Especially:

  • 🐛 Bug reports
  • 🧪 Test cases & validation studies
  • 💡 New use cases

If you're simulating something interesting, we want to hear about it.

GitHub logo showjihyun / Prophet

Prophet

Prophet — 3D social simulation spreading in real time

🔮 Prophet

The wind tunnel for marketing campaigns

Test your campaign on 10,000 AI agents before you spend a dollar on the launch.

GitHub stars License: MIT Version Last commit

Python FastAPI React TypeScript Docker Tests


🚀 Quick Start · ✨ Features · 🎯 Use Cases · 📊 Comparison · 📖 Docs · 🤝 Contributing


git clone https://github.com/showjihyun/prophet.git
cd prophet && docker compose up -d
open http://localhost:5173
Enter fullscreen mode Exit fullscreen mode

That's it. 5 minutes from clone to your first simulation. No API keys required to start — Prophet runs fully locally on a laptop.

💡 Why Prophet?

Focus groups lie — 10 humans in a room cannot tell you how a message spreads through a community. A/B tests are too late — by the time you have data, you're already paying for the launch. Brand-lift studies take 6 weeks, cost $50K, and tell you nothing about why a message failed.

Every discipline that ships things at scale — aerospace, civil engineering, software — gets to simulate…

View on GitHub

Top comments (0)