Ember

Ember is a public AI prediction engine built on a radical premise: an AI that won't show its work isn't worth trusting. Every morning at 7:00 AM EST, three genuinely different AI models—Claude by Anthropic, Grok by xAI, and Gemini by Google—independently call live Polymarket markets before they resolve. They do not consult each other. They do not share notes. When an AI's probability diverges from the real-money crowd by 10 or more points, that divergence gets flagged as a high-conviction signal. Every call is timestamped before the outcome is known. Nothing is edited after the fact. Nothing is deleted. Every wrong call gets a public post-mortem. The record builds in public over 365 days. Accuracy is tracked using Brier scores, a calibration metric that rewards both accuracy and confidence. The model that beats the crowd most consistently across the full year wins. Ember is built for serious bettors, prediction market enthusiasts, and anyone who wants a transparent, verifiable edge in forecasting. It synthesizes data from 20 sources across prediction markets, bookmaker odds, AI research feeds, and real-time sentiment before making any call. Subscribers see the signals at 7:00 AM EST before public release. The edge is speed. The proof is the record.

Three Independent AI Models Daily

Every morning at 7:00 AM EST, Claude, Grok, and Gemini each independently assign probability to live Polymarket markets. They do not consult each other. They do not share data. Claude reasons carefully from first principles, synthesizing prediction markets, bookmaker lines, and AI research feeds. Grok reads live X sentiment for recency and cultural awareness. Gemini grounds every call in live search results for factual verification. When they disagree, that disagreement is logged. Consensus is not the goal. Divergence is the signal.

Divergence Flagging System

When any AI model's probability diverges from the Polymarket real-money crowd by 10 or more points, that divergence is automatically flagged as a high-conviction signal. This is where the edge lives. Either the crowd is wrong or the AI is wrong. The public record shows which. Every flagged divergence is timestamped, locked before the outcome, and scored at resolution. No edits. No deletions. The system tracks every divergence across the full 365-day experiment.

Public 365-Day Accuracy Record

Every call is tracked using Brier scores, a calibration metric that rewards both accuracy and confidence. The model that beats the crowd most consistently across 365 days wins. Nothing is edited after the fact. Every wrong call gets a public post-mortem explaining what went wrong. The record builds in public, completely transparent and verifiable. Anyone can audit the calls, the timestamps, and the outcomes. This is the public proof layer for AI prediction.

Real-Time Data Synthesis Engine

Ember pulls live data from 20 sources before every call. This includes real-money markets like Polymarket, Manifold, and Metaculus (volume-filtered at $10k+), live H2H bookmaker lines from 40+ sportsbooks worldwide, AI research feeds from arXiv cs.AI, Hugging Face Papers, OpenAI Blog, DeepMind Blog, and Astral Codex Ten, plus emerging product signals from Product Hunt, Hacker News, BetaList, GitHub Trending, and Y Combinator launches. Three models reason over all of it simultaneously.

Identifying Market Mispricings Before They Correct

Serious bettors use Ember to spot prediction market mispricings before the crowd adjusts. When an AI model diverges from the real-money crowd by 10+ points, that divergence signals a potential edge. Subscribers see the signal at 7:00 AM EST before public release. The edge is timing. Every call is locked before the outcome, so users know the signal was genuine and not edited after the fact. This turns AI disagreement into actionable betting intelligence.

Benchmarking AI Model Performance

Researchers and AI enthusiasts use Ember's 365-day public record to benchmark how different AI architectures perform at probabilistic forecasting. Claude, Grok, and Gemini each bring different strengths—careful reasoning, real-time sentiment, and factual verification. The Brier score tracking shows which approach is most calibrated over time. This is the first long-term, transparent, real-money benchmark of AI prediction capabilities.

Building a Transparent Decision Record

Organizations and individuals who value transparent decision-making use Ember as a model for how to build a public record of predictions. Every call is timestamped, locked, and scored. Wrong calls get post-mortems. Nothing is hidden. This creates accountability and forces continuous improvement. It is a template for how AI systems should operate when real stakes are involved.

Real-Time Market Sentiment Analysis

Traders and analysts use Ember's divergence signals as a real-time indicator of market sentiment versus AI reasoning. When Grok reads live X sentiment and diverges from the crowd, that reveals whether social media narratives are driving market prices. When Gemini checks facts and diverges, that shows whether the crowd is ignoring verifiable information. This multi-model approach provides a richer picture of market dynamics than any single source.

What makes Ember different from other prediction tools?

Ember is the only prediction engine that runs three genuinely independent AI models daily, logs every divergence from the crowd, and builds a fully transparent 365-day public record. Every call is timestamped before the outcome. Nothing is edited or deleted. Wrong calls get public post-mortems. Accuracy is tracked using Brier scores. This level of transparency and accountability does not exist anywhere else in the prediction market space.

How do I access the signals before they are public?

Subscribers pay $29 per month to see the signals at 7:00 AM EST before public release. The edge is timing. Public release follows shortly after, but subscribers get the divergence signals first. Every call is locked before the outcome, so subscribers know the signal was genuine and not adjusted after the fact. The subscription also unlocks the full divergence gap data behind each call.

How is accuracy measured across the 365-day experiment?

Accuracy is tracked using Brier scores, a standard calibration metric that rewards both accuracy and confidence. A lower Brier score means better calibration. The model that beats the crowd most consistently across the full 365 days wins. Every call is scored at resolution. The results are public and verifiable. No edits. No deletions. The record builds in real time for anyone to audit.

Can I see the reasoning behind each AI model's call?

The system shows which model made each call and the probability assigned. The divergence from the crowd is displayed. However, the full reasoning chain for each model is not currently published. The focus is on the output, the divergence signal, and the accuracy record. Subscribers see the full divergence gap data. The public record shows the calls, timestamps, and outcomes for complete transparency.