What question did this study set out to answer?

The central aim is to design an architecture that allows users to define rules for AI agents while ensuring trustworthiness in results.

April 16, 2026Open Access

Engine, Rules, and Canon: An Architecture for User-Owned, Continuously Iterating AI Agents

View Full Paper

Ask AI

Bookmark

Authors

This framework demonstrates user-owned AI agents' results in various domains, highlighting their adaptability and reliability.

The central aim is to design an architecture that allows users to define rules for AI agents while ensuring trustworthiness in results.
Developed a domain-agnostic engine for applying user-owned rules.
Implemented a continuously iterating kernel optimizing a trustworthiness metric called NetTrust.
Introduced a four-primitive witness stack for verdict categorization.
Established a matched-paper evaluation protocol to test a Reviewer-Correlation Ceiling Hypothesis.
Introduced systems for making disposable results from user-defined rules.
Showed that the architecture maximizes trustworthiness through self-verification mechanisms.
Validated that the reviewer ensembles achieve substantial agreement under certain conditions.

Andrew Salvo (2026) studied this question.

View Full Paper

Ask AI

Bookmark