soft-predict-future

Ask any future-oriented question such as "Will crypto replace banks?" or "Who will win the AI race?" and the engine runs a 9-step pipeline. It collects web signals, scores them, and outputs Probable, Plausible, Possible, and Preferable scenarios with confidence scores.

No, a year is not required. The engine infers a reasonable horizon from the question type: 3-10 years for competitive races, 5-15 years for technology adoption, 10-20 years for geopolitical shifts, and 2-5 years for company survival questions.

Yes, web search is mandatory. The pipeline runs exactly 6 searches covering current state, growth, barriers, policy, technology enablers, and historical precedent, collecting at least 18 signals before proceeding with scoring.

Soft predict uses Claude's native reasoning with web search for fast, structurally correct outputs. Hard predict future is a separate deterministic mode with auditable scoring, invoked by saying "Run hard predict future" followed by your question.

Questions fail validation if the entity is fictional, the domain is purely philosophical, the timescale exceeds 30 years, no real-world evidence could exist, or the question is too vague to produce distinct scenarios. The engine explains which of the 5 rules failed.

Signal scores are adjusted with regional multiplier tables for India, USA, Europe, China, or Global default. For example, India amplifies Technological/Operational signals due to UPI/DPI infrastructure, while Europe amplifies Environmental/Strategic signals due to regulatory leadership.