Build domain-expert AI from messy historical data

Lightning Rod turns raw documents and public sources into verified training sets and compact domain experts — without hand-labeling.

Real-world data has timestamps, not clean labels.

Generate verified training data from real-world outcomes.

DATASET Policy Forecast |Example 457 / 5,000

Question
Will the Trump administration impose 25% tariffs on all goods from Canada by March 1, 2025?

Question Source
New York Times Jan 27, 2025

Label
Yes.

Type
binary

Confidence
0.92

Label Source
Reuters Feb 1, 2025

DATASET Medical QA |Example 117 / 4,200

Question
What is the mechanism by which beta-blockers reduce mortality in heart failure patients?

Source
Harrison's Internal Medicine Chapter 257 — Heart Failure: Management, pp. 1762–1769

Answer
Beta-blockers block β₁-adrenergic receptors, reducing heart rate and myocardial oxygen demand, allowing reverse remodeling and improved systolic function.

Type
free response

Confidence
0.94

Excerpt
Beta-blocker therapy reverses adverse LV remodeling by attenuating the cardiotoxic effects of sustained adrenergic activation.

DATASET Supply Chain Disruption |Example 814 / 3,200

Question
What will the Global Supply Chain Pressure Index (GSCPI) value be for March 2025?

Seed
≡ gscpi_historical.csv

Label
1.84

Type
continuous

Context
Financial Times Red Sea Disruptions Push Supply Chain Stress to 18-Month High

DATASET Portfolio Company Risk |Example 33 / 2,400

Question
Will ProServ Health's largest payer contract be renewed before its June 2025 expiration?

Question Source
ProServ Health — Q4 2024 Operating Review Feb 3, 2025

Label
Yes.

Type
binary

Confidence
0.98

Label Source
ProServ Health — Q2 2025 Board Presentation Jul 18, 2025

Prompt to AI

Describe what you want. Our agent handles the rest.

I want to predict the likelihood of geopolitical events using news data.

Got it. I'll pull from Reuters and AP News — about 18 months of coverage. Does that work?

Yes, go ahead.

Gathering sources now. You can track progress on the right.

The agent shows its reasoning at every step — you confirm before it commits.

Used to train frontier-beating models.

#1 on ProphetArena Sports

Foresight-32B ranked #1, ahead of GPT-5.2 and Gemini 3 Pro.

Top 5 on ForecastBench

Outperformed Gemini 3 Pro, Claude Sonnet 4.5, and o3 on the Forecasting Research Institute benchmark.

Cutting Edge Research

Beating frontier models using our novel Future-as-Label methodology.

Simple, powerful API

Generate verified datasets in a few lines of code. Our SDK handles the complexity.

  • Grounded in real outcomes and source documents
  • Bootstrap with public feeds: news, SEC filings, Wikipedia
  • Full provenance with citations and source docs

Example Code

from lightningrod import Pipeline

pipeline = Pipeline([
    NewsSeedGenerator(query="AI regulation"),
    ForwardLookingQuestionGenerator(
        instructions="Generate questions about future AI regulations and rulings"
    ),
    WebSearchLabeler()
])

dataset = pipeline.run(n_samples=100)

Train AI experts for any domain.

See how Lightning Rod turns your sources into verified training data in minutes.