Introducing /forecast: Structured Predictions for AI Agents

Predict What's Coming Next

Your agent can now predict what's coming next — not just report what already happened. The new /forecast endpoint takes a topic, analyzes thousands of verified briefs for emerging patterns, and returns structured predictions with confidence scores, evidence chains, and explicit falsification criteria.

This is forward-looking intelligence as a service. Ask any question about what's likely to happen, and get back machine-readable predictions your agent can act on.

How It Works

Send a topic to POST /api/v1/forecast. Polaris searches its brief corpus for relevant context, identifies signals and trend lines, and generates structured predictions.

Two depth tiers:

What You Get Back

The response is fully structured for programmatic consumption:

{
  "executive_summary": "Semiconductor supply chain constraints are...",
  "trend_direction": "tightening",
  "predictions": [
    {
      "prediction": "TSMC will announce capacity expansion in Q2 2026",
      "confidence": 0.72,
      "timeframe": "60-90 days",
      "evidence": [
        "Brief PR-4821: TSMC board approved $28B capex...",
        "Brief PR-4956: Arizona fab construction ahead of schedule..."
      ],
      "invalidated_if": "TSMC delays Arizona fab or cuts capex guidance"
    }
  ],
  "key_signals_to_watch": [
    "TSMC monthly revenue reports",
    "US CHIPS Act disbursement timeline"
  ],
  "wildcards": [
    "Escalation in Taiwan Strait tensions could override all supply forecasts"
  ]
}

Every prediction includes an invalidated_if field — explicit conditions that would break the prediction. Your agent can monitor these as ongoing checks.

Python SDK

from polaris_news import PolarisClient

client = PolarisClient()  # uses POLARIS_API_KEY env var

forecast = client.forecast("semiconductor supply chain", depth="deep")

print(f"Direction: {forecast.trend_direction}")
print(f"Summary: {forecast.executive_summary}")

for p in forecast.predictions:
    print(f"\n[{p.confidence:.0%}] {p.prediction}")
    print(f"  Timeframe: {p.timeframe}")
    print(f"  Invalidated if: {p.invalidated_if}")

print(f"\nWildcards: {forecast.wildcards}")
print(f"Signals to watch: {forecast.key_signals_to_watch}")

Use Cases

Trading signal validation

Before your agent acts on a market signal, check what the broader intelligence landscape predicts. Cross-reference your signal against Polaris forecasts to reduce false positives.

Risk assessment

Supply chain risk, regulatory risk, geopolitical risk — get structured predictions with evidence so your risk models have forward-looking context, not just historical data.

Competitive intelligence

Track what's likely to happen next in your competitive landscape. Monitor wildcards that could disrupt your market.