Every AI claims it can predict things; almost none can prove the prediction came first. NeuPortal is a public forecasting experiment that does. Our AI's probabilities for sports, crypto and prediction markets are locked before each event, hashed into Bitcoin via OpenTimestamps so nothing can be backdated, benchmarked against the market price at the same instant, and Brier-scored after the result. A live scoreboard shows every call — including the nights the market beats us. That honesty is the product. Educational transparency project, not financial advice.
Forecasts locked before every event and hashed into Bitcoin via OpenTimestamps — no backdating, ever
Every call Brier-scored in public against the prediction-market price frozen at the same instant
Live scoreboard shows the full record — wins and losses alike
Open, verifiable dataset: Zenodo DOI + GitHub proofs repo with a one-command verify script
Covers 2026 World Cup, crypto and macro forecasts
Check whether an "AI predicted X" claim was actually made before the event
Learn forecasting literacy — calibration, Brier scoring, base rates — from a live example
Benchmark an AI model's probabilistic accuracy against a market baseline

Founder here. We built NeuPortal because "our AI predicted X" is the most unfalsifiable sentence on the internet — any track record shown after the event could be edited or cherry-picked. So we built the audit into the pipeline instead of asking anyone to trust a screenshot: every forecast is locked before the event, hashed into Bitcoin via OpenTimestamps, benchmarked against the prediction-market price frozen at the same instant, and Brier-scored in public. The honest part — right now the market is beating our model, and we publish that; the losses are the point. Everything is verifiable: open dataset + a one-command proof script. Happy to answer anything about the timestamping pipeline or the scoring.

Founder here. We built NeuPortal because "our AI predicted X" is the most unfalsifiable sentence on the internet — any track record shown after the event could be edited or cherry-picked. So we built the audit into the pipeline instead of asking anyone to trust a screenshot: every forecast is locked before the event, hashed into Bitcoin via OpenTimestamps, benchmarked against the prediction-market price frozen at the same instant, and Brier-scored in public. The honest part — right now the market is beating our model, and we publish that; the losses are the point. Everything is verifiable: open dataset + a one-command proof script. Happy to answer anything about the timestamping pipeline or the scoring.
Find your next favorite product or submit your own. Made by @FalakDigital.
Copyright ©2025. All Rights Reserved