Kalpae

What it is:

• Kalpae is an automated forecasting system that tracks how likely future events are to happen by continuously researching the web and updating probabilities based on real evidence. Instead of relying on guesses or betting markets, Kalpae automatically gathers information, extracts signals using AI, and updates forecasts using transparent rules. Each probability change is recorded along with the evidence that caused it, creating a verifiable history of how beliefs about future events evolve.
  How it works:
• Kalpae periodically researches the internet using Chainlink’s Computation Runtime Environment, which runs automated workflows to discover relevant articles and documents from trusted sources. These documents are analyzed using AI to extract structured signals such as relevance, confidence, and whether the evidence supports or contradicts the event. A deterministic belief engine then calculates how the probability should change based on these signals. Each update is committed to the Solana blockchain, creating an immutable record that links the previous probability, the new probability, and a cryptographic hash of the supporting evidence.
  What problem it solves:
• Predictions today are often opaque and difficult to verify. People can change their beliefs without explanation, claim they predicted something correctly without proof, or present probabilities without showing the evidence behind them. Kalpae solves this by creating a transparent system where forecasts evolve only when new evidence appears, and every update is recorded with a clear explanation and an immutable record. This makes forecasting explainable, auditable, and trustworthy.

How is it built?

Kalpae is built using a hybrid architecture that combines modern web technologies, automated off-chain computation, AI-based signal extraction, and blockchain verification to create a transparent and auditable forecasting system.

The user interface and application layer are built with Next.js and TypeScript, using Tailwind CSS for styling and tRPC with React Query for fully type-safe communication between the frontend and backend. Users interact with the system through a web interface and authenticate using a Solana wallet, which provides secure identity without traditional passwords.

The backend uses PostgreSQL as the main database and Prisma as the ORM. This database stores questions, evidence documents discovered from the web, AI-extracted signals, and probability updates that form the belief timeline.

Automated research and processing are handled off-chain by Chainlink’s Computation Runtime Environment (CRE). CRE runs scheduled workflows that search the web for relevant articles, fetch and process documents, and send them to AI models for analysis. The AI models extract structured signals such as relevance, confidence, and whether the information supports or contradicts the predicted event.

These signals are then processed by a deterministic belief engine, which calculates small probability updates using transparent rules. Because the belief engine is deterministic, the same inputs always produce the same probability changes.

To ensure the system is verifiable, the results of each probability update are committed to a Solana program built with the Anchor framework. Instead of storing full data on-chain, the program records cryptographic commitments that link probability updates to the underlying evidence and computation version. This creates an immutable audit trail that proves the forecasting history has not been tampered with.

Together, this architecture allows Kalpae to combine automated research, AI signal extraction, deterministic probability updates, and blockchain verification to create a transparent and trustworthy forecasting system.