> ## 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 Signals

> Multi-agent predictions with sourced reasoning. Your alpha feed.

# When the AI disagrees with the market

A signal fires when Supermission's AI system calculates a probability that meaningfully diverges from the current market price. That divergence is your potential edge.

Signals aren't tips. They're reasoned analyses with sources, confidence scores, and full transparency into how the AI reached its conclusion.

## Three signal types

<CardGroup cols={3}>
  <Card title="Sentiment" icon="brain">
    AI's independent probability estimate based on news, social data, and fundamentals. Fires when AI confidence diverges from market price.
  </Card>

  <Card title="Momentum" icon="arrow-trend-up">
    Price action confirms what sentiment already predicted. The market is moving toward the AI's estimate. Trend-following signal.
  </Card>

  <Card title="Spot Price" icon="bitcoin">
    Crypto-specific. Triggers when a token crosses a threshold relevant to a prediction market (e.g., BTC above `$100`k triggers related market signals).
  </Card>
</CardGroup>

## How signals are generated

This isn't a single model guessing. It's a 4-agent system:

<Steps>
  <Step title="Independent analysis">
    Four AI agents — Bull, Bear, Contrarian, and Quant — independently analyze the market. Each has different biases and data priorities. They don't see each other's work.
  </Step>

  <Step title="Evidence gathering">
    Each agent pulls from news feeds, social sentiment, on-chain data, and historical patterns. Every claim is sourced.
  </Step>

  <Step title="Debate synthesis">
    A judge agent reviews all four analyses, weighs the arguments, and produces a final probability estimate with confidence score.
  </Step>

  <Step title="Signal emission">
    If the synthesized probability diverges meaningfully from market price, a signal fires with direction (YES/NO), confidence (0-100), and the full reasoning chain.
  </Step>
</Steps>

## Confidence scores

Every signal carries a confidence score from 0 to 100:

| Range  | Meaning                                                               |
| ------ | --------------------------------------------------------------------- |
| 80-100 | Strong conviction. Agents largely agree. Evidence is clear.           |
| 60-79  | Moderate conviction. Some disagreement between agents.                |
| 40-59  | Low conviction. Significant agent disagreement. Proceed with caution. |

## Filtering signals

The signal feed can be filtered by:

* **Category** — Politics, Crypto, Sports, Finance, etc.
* **Confidence** — Minimum threshold slider
* **Volume** — Filter out thin markets where execution is harder

<Tip>
  High confidence + high volume + large divergence from market price = the strongest setups. That's the trifecta. Most traders filter to 70+ confidence and sort by edge size.
</Tip>

<Warning>
  Signals are analysis, not financial advice. The AI is good, but markets are adversarial. Always understand the reasoning before trading on a signal — that's why we show the full chain.
</Warning>