How to Request an AI Prediction
Step-by-step guide to requesting an AI-generated prediction on Eroteme.
Requesting an AI prediction takes less than a minute. Here's how.
Step-by-Step Guide
1. Open the AI Prediction Modal
From any prediction page on Eroteme, click Request AI Analysis. This opens the prediction request modal.
2. Choose an Event
Browse or search for an active prediction market. You can filter by category — sports, politics, crypto, entertainment, and more.
3. Choose Your Tier
| Tier | What You Get | Cost |
|---|---|---|
| Standard | Single AI model analysis (Claude, GPT, Gemini, or Grok) | $1 USDC |
| Eroteme | Full 8-agent pipeline with meta-judge synthesis | $5 USDC |
Standard gives you a quick, single-perspective analysis from one AI model. Good for straightforward markets.
Eroteme runs the full pipeline — 4 research agents gather real-time data, 4 analysis agents argue bull/bear/evidence/contrarian perspectives, and a meta-judge synthesises everything into a calibrated probability. Best for high-stakes or complex markets.
4. Review the Fee and Confirm
The modal shows the analysis fee in USDC. Review and confirm. Remember: if the AI prediction turns out to be wrong, you can claim a refund.
5. Watch the Pipeline Run
Once confirmed, the AI agents begin analysing the event. For Eroteme tier, you can watch the two-layer pipeline progress:
Research Layer (parallel):
- Perplexity gathering latest news
- Grok analysing social sentiment
- Gemini pulling data and statistics
- Claude computing base rates
Analysis Layer (parallel, after research completes):
- Claude building bull case
- ChatGPT building bear case
- Gemini synthesising evidence
- Grok finding contrarian angles
Meta-Judge (final):
- Claude with Extended Thinking synthesises all 8 outputs
For Standard tier, a single model runs and returns its analysis.
6. Review the Output
When the pipeline completes, you see the full results:
- Probability — A calibrated percentage (0–100%) for the outcome
- Confidence interval — Uncertainty bounds around the probability
- Disagreement score — How much the agents diverge (Eroteme only)
- Per-agent breakdown — Each agent's probability, confidence, and reasoning (Eroteme only)
- Key factors — The main drivers behind the prediction
- What could be wrong — The biggest risk to the estimate
See Understanding the AI Output for a detailed explanation of each component.
7. Publish as a Live Prediction
Once you're satisfied with the analysis, publish it as a live prediction on the platform. Other users can then view the analysis, see the per-agent breakdown, and bet on the outcome.
Publishing an AI prediction creates a standard prediction on the platform. Other users can bet for or against it just like any user-created prediction.
Eroteme