Registry for Genuine Information, Notarised Authentication, and Legitimate Documentation
When AI agents guess, people lose money. REGINALD is the global registry where provenanced, machine-readable documentation lives and is queryable by any AI system worldwide.
AI inference systems generate wrong answers to questions that already have documented answers. This happens because documentation was designed for humans reading web pages — not for machines parsing structured data.
The £200,000 pricing error
A river cruise priced at £2,030 was reported by an AI agent as £203,000. European number formatting, no Schema.org PriceSpecification, no structured price data. The couple never booked. The cruise company never knew why it lost the sale.
The 40,000-token waste
A developer asked an AI coding assistant about a documented authentication command. The command existed. The AI denied it. The troubleshooting cascade consumed over 40,000 tokens across two AI systems. The correct answer required 200 tokens. A 200x cost multiplier — on a single question.
REGINALD and COGs exist to eliminate this waste at the source. A single API call replaces the cycle of web search, page parsing, prose extraction, and speculative synthesis.
| Without REGINALD | With REGINALD |
|---|---|
| Web scraping (unreliable, expensive) | Structured registry (reliable, efficient) |
| No verification of source accuracy | Cryptographic attestation (Ed25519) |
| No accountability for wrong information | Named maintainer + SLA + review cycle |
| Content goes stale silently | Expiry dates + update triggers |
| Each AI platform scrapes differently | One format, universal access |
| Inference from prose (error-prone) | Structured metadata — YAML, JSON-LD, Schema.org (precise) |