How Q works

Q runs a fixed loop. Scored outcomes from published forecasts feed the next ones.

Select markets. The prediction markets worth forecasting.
Identify sources. The people and publications that move each market.
Collect data. Articles, posts, and prices, continuously.
Forecast. The pipeline turns that stream into one probability per market.
Score. Every forecast is scored against the outcome.
Build strategies. Correlations in the data become products: direct market entries, and cross-signals such as the Hawk & Dove index against oil.
Fine-tune. Scored outcomes feed LoRA fine-tuning for the next model version.

What the API serves

The API is the output of that loop, in four objects:

  • Assets — the underlying company, commodity, or cryptoasset.
  • Markets — the prediction-market contracts about it, with live venue odds.
  • Forecasts — Q's probability on a market, with its bands and reasoning.
  • Signals — the calls Q publishes, with frozen entries and a live record.

Alongside them: the sources behind each forecast, the Hawk & Dove index, portfolio intelligence, and the record of everything published.

How objects link

An Asset has Markets. A Market has Forecasts. A Signal sits on a Forecast, on a Market. Detail responses can include a relationships envelope with the object's graph neighborhood: flat identity references, up to 50 per category.

Each reference keeps the exact relationship, direction (incoming or outgoing), and via (direct, market, or asset). Check relationships.truncated before treating a neighborhood as complete, and follow a reference to its canonical detail route when you need more than the identity. Workflows covers this in the detail an agent needs.