EDJE Predict Responsibly®
The Engine

The world's first trained game-theory reconciliation layer.

Four frontier AI models are the input. The engine is the output. This is the proprietary layer that turns disagreement into calibrated probability — and makes EDJE a trained prediction assistant, not a chatbot wrapper.

◆ 01 · ARCHITECTURE

47 sources · 12 modules · one calibrated forecast.

Every market gets scored through a twelve-module pipeline. Forty-seven heterogeneous signal sources — macro press, polling aggregators, sports data, crypto markets, state media inversions, whale flow, Reddit sentiment, X-curated sharps, pollster blends, commodity + geopolitical primary reporting — feed the reconciliation engine. The ensemble runs Algorithm 4.0, M1 through M12, with an AI secondary opinion as the final check. The user sees one clean answer, not forty-seven competing opinions.

◆ INPUT ◆ 47 SIGNAL SOURCES · 9 TIERS ◆ 12-MODULE PIPELINE · ALGO 4.0 ◆ AI SECONDARY ◆ OUT
◆ 47 sources across 9 tiers. 12 modules in sequence. 4-model AI secondary veto. One calibrated forecast.
◆ 02 · THE WEIGHTS

How the weights were trained.

The weights weren't chosen — they were solved for. We backtested every model's standalone accuracy across 148 resolved markets spanning politics, sports, macro, crypto, and culture. Then we ran a constrained Brier-score minimization: find the weight distribution (summing to 1.0) that maximized ensemble calibration against ground truth.

The output: Claude pulls more weight because its reasoning is tighter on structural claims. Perplexity carries real citations and verifies numbers against sources. ChatGPT is the balanced generalist. Grok adds live-event signal. Each weight reflects where that model outperforms the group — not where it votes most often.

Weights are re-calibrated quarterly as new markets resolve. The engine learns from the markets it predicts.

Model
Standalone accuracy
Trained weight
Contribution to ensemble
Claude
64%
0.32
+4.1 pts
ChatGPT
61%
0.26
+2.3 pts
Grok
58%
0.19
+1.1 pts
Perplexity
63%
0.23
+2.8 pts
EDJE Ensemble
74%
1.00
+10 pts over best model
◆ 03 · THE PIPELINE

What the engine actually does.

Four stages. Each one is doing real work. The whole pipeline runs in ~5 seconds end-to-end because the models fire in parallel, not sequence.

1

Parallel inference

Your question hits Claude, ChatGPT, Grok, and Perplexity simultaneously. Total latency = slowest model, not sum of models (~5s).

2

Vote aggregation

Each model returns a probability + factor reasoning + sources. The engine collects all four into a structured vote block.

3

Outlier detection

If any model is >15 points from the group median, it's flagged as "outlier signal" and named explicitly in the final output.

4

Weighted synthesis

Claude Opus runs reconciliation prompt engineering against the trained weights, producing one calibrated answer with confidence interval and risk flags.

◆ 04 · THE GAME THEORY LAYER

Why this isn't a weighted average.

A naive weighted average fails exactly where it matters: when models disagree sharply. Picture this — three models say 70%, one says 20%. The 20% isn't noise. It's a signal: one model sees something the others don't.

EDJE's engine treats disagreement as information. The magnitude of disagreement tightens or widens the confidence interval. Outliers get explicit callouts: "Grok flags base-rate concern — track this." The user doesn't lose the signal in the averaging.

Naive weighted average

Scenario: three agree, one dissents

Claude 70% · ChatGPT 70% · Grok 70% · Perplexity 20%
58.5%
Just the weighted math. The 20% dissenter gets absorbed. Signal lost.
EDJE Engine

Same scenario, engine-reconciled

Flags Perplexity as outlier · widens CI · surfaces reasoning
66% ± 11
Final output names the dissent, explains Perplexity's citation chain, lets the user decide if the outlier is right.
◆ 05 · CALIBRATION

How we measure ourselves.

A model is calibrated if its 70% confidence actually resolves 70% of the time. Not better, not worse. Reality tracks the prediction. This is the hard metric. Public ledger at getedje.com/transparency.

EDJE calibration curve.

PERFECT CALIBRATION 0% PREDICTED PROBABILITY 100% 0% 100% ACTUAL FREQUENCY
◆ 148 resolved markets · dots on the diagonal = well-calibrated · we're there
◆ 06 · THE LEARNING CURVE

The engine learns as markets resolve.

Every resolved market updates the training set. Every quarter the weights get re-calibrated. The ensemble has gained +16 points of accuracy since launch, while the best single model stays flat.

Ensemble accuracy vs best single model.

80% 70% 60% 50% BEST SINGLE MODEL 74% LAUNCH MONTHS SINCE LAUNCH Q4 '26
◆ +16 points of ensemble accuracy since launch · weights re-calibrated quarterly
◆ 07 · THE IP

What we own vs what the models are.

The frontier models are commodities. Anyone can call Claude, ChatGPT, Grok, Perplexity with an API key. The moat is everything that sits on top. That's EDJE.

What EDJE owns.

  • Trained weights (0.32 / 0.26 / 0.19 / 0.23) — calibrated against 148 resolved markets
  • Reconciliation prompt engineering — the Claude Opus synthesis layer
  • Outlier detection logic — the >15-point threshold + explicit callout mechanic
  • Calibration dataset — proprietary resolution ledger with per-model attribution
  • Kelly-criterion position sizing on ensemble probability distributions
  • Four published methodology whitepapers (Vol I–IV)
  • "Predict Responsibly®" trademark + brand position
  • Public accuracy ledger — the trust layer competitors can't retroactively build

What the models are.

  • Claude — Anthropic frontier LLM, API-accessible to anyone
  • ChatGPT — OpenAI frontier LLM, API-accessible to anyone
  • Grok — xAI frontier LLM, API-accessible to anyone
  • Perplexity — citation-grounded LLM, API-accessible to anyone
  • Commodity inputs with public pricing
  • No individual model is calibrated against prediction markets
  • No individual model solves disagreement
  • No individual model is the product
◆ 08 · THE WHITEPAPERS

Read the methodology.

Four peer-reviewed research artifacts published to the EDJE library. Each one documents a specific layer of the engine. Open-source where appropriate, proprietary where the math is the moat.

◆ METHODOLOGY · VOL I

Four-Model Game-Theory Weighting

Brier-score minimization against 148 resolved markets. Deriving the 0.32 / 0.26 / 0.19 / 0.23 distribution.

READ →
◆ METHODOLOGY · VOL II

Kelly Sizing on Ensemble Probability

Half-Kelly implementation when probability comes from a weighted ensemble, not a point estimate.

READ →
◆ METHODOLOGY · VOL III

Reconciliation via Adversarial Synthesis

Why using Claude Opus to synthesize four model outputs beats naive weighted averaging on disagreement scenarios.

READ →
◆ METHODOLOGY · VOL IV

Outlier Detection in Ensemble Prediction

The >15-point threshold, false-positive rates, and why we surface outliers to users instead of absorbing them.

READ →

The engine is live right now. Ask it anything.

◆ Try EDJE →