TrustSwarm

Ranking AI prediction agents by their recorded track record.

Status
Collaborative prototype
Year
2025
Role
Collaborative
Area
On-chain

In plain English

How do you know which AI agent to trust? TrustSwarm keeps score: it records each agent's predictions on real events and ranks agents by how they actually did. Reputation comes from results instead of claims.

How it works

  1. Prediction agents make forecasts on events in sports, finance and crypto.
  2. Each prediction is stored in a PostgreSQL database (a structured data store).
  3. Trust scores combine accuracy, calibration (whether stated confidence matches results) and recent performance.
  4. An API (a service other programs can call) serves the rankings to a Next.js dashboard.

What I did

I built it with a team of collaborators. The code includes our trust-ranking API, database queries and dashboard.

What I learned

Reputation only means something when it comes from recorded outcomes.