> ## Documentation Index
> Fetch the complete documentation index at: https://docs.supermisson.fun/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Agents Overview

> Multi-agent intelligence for prediction markets. 4 independent agents debate every market. You see the reasoning, not just the number.

# AI that shows its work

Every prediction market tool gives you a probability. Supermission gives you the **argument**.

Our AI system isn't a single model making guesses. It's four independent agents with different analytical biases — Bull, Bear, Contrarian, and Quant — that analyze every market, debate the evidence, and produce a synthesized view with full transparency.

## The philosophy

**Don't delegate understanding.** Every signal shows its thesis. Every edge card shows the full reasoning chain. Every War Room analysis shows the agent debate. You're not trusting a black box — you're evaluating arguments from four distinct perspectives and deciding for yourself.

## The 4-agent system

| Agent          | Role                                       | Analytical bias | Data priorities                                                                        |
| -------------- | ------------------------------------------ | --------------- | -------------------------------------------------------------------------------------- |
| **Bull**       | Narrative and source credibility analysis  | Qualitative     | Source credibility, political dynamics, institutional incentives, social context       |
| **Bear**       | Microstructure and price-in analysis       | Structural      | Whether news is already priced in, trading volume, liquidity, bid-ask spread dynamics  |
| **Contrarian** | Asks "what if everyone is wrong?"          | Skeptical       | Crowded trades, false consensus, scenarios the market isn't pricing, regime changes    |
| **Quant**      | Base-rate anchoring with statistical rigor | Quantitative    | Historical frequency, current market prices vs base rates, time remaining, calibration |

A **Judge** agent synthesizes all four analyses into a final probability + confidence score, weighing argument quality, evidence strength, and historical calibration.

## AI features across the terminal

<CardGroup cols={2}>
  <Card title="AI Signals" icon="signal" href="/ai/signals">
    Real-time alerts when the AI's probability diverges from market price. Three types: Sentiment (research-driven edge), Momentum (price confirms AI), Spot Price (crypto threshold triggers). Confidence-scored 0-100.
  </Card>

  <Card title="Signal Feed" icon="list" href="/ai/signal-feed">
    Dedicated filterable feed of all active signals. Filter by type, category, confidence minimum. Sort by confidence, volume, ending soon, or price change. Cursor-based infinite pagination.
  </Card>

  <Card title="AI War Room" icon="users" href="/ai/war-room">
    The flagship feature. 4-agent debate with Edge Calculator, Conviction Meter (with audio chime at extreme levels), agent cards with reasoning, Key Doubt from the Contrarian, and Accuracy Scoreboard.
  </Card>

  <Card title="Oracle Agent" icon="message-bot" href="/ai/oracle-agent">
    Chat-based trading assistant in the Markets terminal. Search markets, analyze positions, place trades, sell positions, manage portfolio — all through natural language with SSE streaming responses.
  </Card>

  <Card title="Whale Intelligence" icon="fish" href="/ai/whale-intelligence">
    Top 20 holders per market with position sizes, outcome badges, and aggregate whale sentiment. Surfaces "Smart money leans YES/NO" when whale positions lean 55%+ in one direction.
  </Card>

  <Card title="Alpha Scanner" icon="crosshairs" href="/ai/alpha-scanner">
    Automated edge detection running every 30 minutes. Compares AI probability vs market price, calculates EV and Kelly criterion sizing, classifies edges as high conviction / moderate / speculative.
  </Card>

  <Card title="Auto-Pilot" icon="robot" href="/ai/auto-pilot">
    Autonomous trade execution. Set budget, risk tolerance, and signal preferences. The system executes when high-conviction opportunities appear. Full audit trail for every decision (including skips).
  </Card>
</CardGroup>

## Signal strength hierarchy

The strongest trade setups combine multiple AI layers:

<Steps>
  <Step title="AI Signal fires">
    The multi-agent system finds a meaningful divergence between its probability and the market price. Confidence score is 70+ and agents largely agree.
  </Step>

  <Step title="War Room confirms">
    Edge Calculator shows significant mispricing. Conviction Meter is high (70%+). Agent Debate Panel shows 3/4 or 4/4 agents aligned.
  </Step>

  <Step title="Whales align">
    Top Holders panel shows large positions leaning in the same direction as the AI's thesis. Smart money sentiment crosses 55% threshold.
  </Step>

  <Step title="Execute with confidence">
    When all three layers agree — AI signal + War Room conviction + whale alignment — that's as close to a high-probability trade as prediction markets offer.
  </Step>
</Steps>

## Market tiering

Not all markets get the same AI treatment. Analysis depth scales with market importance:

| Tier       | Criteria                    | Analysis                                 | Update frequency |
| ---------- | --------------------------- | ---------------------------------------- | ---------------- |
| **Tier 1** | Trending OR volume > `$50K` | Full 4-agent debate with judge synthesis | Every 15 minutes |
| **Tier 2** | Volume > `$5K`              | 2-agent quick analysis (Bull + Bear)     | Every 30 minutes |
| **Tier 3** | Everything else             | Single-model sentiment only              | Every 60 minutes |

Markets also get **event-driven** re-analysis when matched news arrives, regardless of tier.

<Warning>
  AI signals are analysis, not financial advice. The multi-agent system improves signal quality but cannot predict the future. Markets are adversarial. Always understand the reasoning chain before acting on a signal — that's why we show the full debate.
</Warning>