Receipt Schema (v1)

Atomic, append-only, verifiable. Each receipt is a cryptographically hashed record of frontier progress.

Hash formula: sha256(url | published_at | title)

type Receipt = {
  id: string;                 // uuid
  source: string;             // host (e.g., "openai.com")
  title: string;
  url: string;                // canonical source link
  published_at: string;       // ISO date/time (UTC)
  frontier?: "AI" | "Energy" | "Biotech" | "Crypto" | "Defense" | "Robotics" | "Frontier Founders";
  hash: string;               // sha256(url|published_at|title)
  summary?: string;
  visible: boolean;
  cid?: string;               // analysis artifact id
  novelty_score?: number;     // 0-1, quality signal
  impact_score?: number;      // 0-1, quality signal
  created_at: string;
};

Design Principles

  • •Immutable: We never overwrite receipts. Corrections are expressed as new receipts (nullifications).
  • •Append-only: Receipts form an audit trail. Deletions are logical, not physical.
  • •Deterministic: Same inputs always produce the same hash. Enables deduplication and verification.
  • •Content-addressed intelligence: CID links to analysis JSON stored in Supabase Storage.
💡 Why this matters

Verified receipts are the foundation for trustworthy AI reasoning. When every claim traces back to a hashed, timestamped source, models become auditable and hallucinations become detectable.