How It Works
The 6-agent debate engine, scoring system, and everything you need to understand your scan results.
The 6 Agents
VerdictSwarm deploys 6 independent AI agents, each analyzing a different dimension of risk. No single model sees the whole picture — that's the point.
| Agent | What It Analyzes | Powered By |
|---|---|---|
| 📊 Technician | On-chain data: holders, liquidity, concentration, transaction patterns | GPT-4o Mini (OpenAI) |
| 🔒 Security | Contract safety: mint/freeze authority, LP locks, honeypot detection, audit status | Claude 3 Haiku (Anthropic) |
| 💰 Tokenomics | Supply distribution, unlock schedules, FDV vs market cap, dilution risk | Gemini 2.5 Flash (Google) |
| 🐦 Social | X/Twitter sentiment, whale activity, fake engagement, breaking news | Grok 3 (xAI) |
| 🌍 Macro | Market conditions, BTC dominance, alt season signals, sector rotation | Grok (xAI) |
| 😈 Devil's Advocate | Adversarial challenge — finds the worst-case scenario for every token | Gemini 2.5 Pro (Google) |
How the Debate Works
Independent Scoring
Each agent scores the token independently on a 0–10 scale before seeing any other agent's result. No groupthink.
Automatic Debate Trigger
If agents disagree significantly, adversarial debate rounds trigger automatically — no manual escalation needed.
4-Phase Debate Loop
Challenge → Defense → Rebuttal → Resolution. Agents must justify position changes with evidence.
Weighted Consensus
Final score is a weighted consensus of all agents after debate adjustment. Some debates end in Split Verdict — both positions are shown to you.
Why debate? A single AI model can be confidently wrong. When six models running on different architectures and training data all independently agree, that consensus is meaningfully stronger than any individual verdict.
Score & Grade Scale
| Score | Grade | Risk Level | What It Means |
|---|---|---|---|
| 80–100 | A / A+ | LOW | Generally safe — standard risk awareness applies |
| 60–79 | B / C | MEDIUM | Caution — notable risks identified, reduce position size |
| 40–59 | D | HIGH | Significant concerns — avoid unless you've done deep manual research |
| 0–39 | F | CRITICAL | Likely scam or extreme risk — do not trade |
Common Questions
😈 Why does the Devil's Advocate always score low?
That's its job. The Devil's Advocate is designed to find the worst-case scenario so the other agents must defend their positions. A token that survives the Devil's Advocate is stronger for it — its low score is the pressure that makes the rest of the swarm justify their optimism.
📉 Why did a popular token get a low score?
The swarm reads the chain, not the hype. If 70% of supply is in 10 wallets, that's a risk regardless of brand name, Twitter following, or how many influencers have mentioned it.
⚠️ Is this financial advice?
No. VerdictSwarm provides automated risk assessments based on on-chain and market data. Results are for informational purposes only. Always do your own research (DYOR) before making any financial decision.
🕐 How fresh is the data?
Scans use live on-chain data and real-time X/Twitter search at the moment of the scan. Cached results are up to 1 hour old — the timestamp is shown on every report.