# Claim boundaries

## Supported as architecture description

- The bilateral loop separates Ouroboros technical authority from Codex adversarial review.
- The release describes memory, evidence, topology, authority, receipts, checkpoints, and process state as parts of that external side.

## Supported by private evidence

- Ouroboros stores consequential system-level intelligence outside the model.
- External state survives individual model calls and affects later work.
- Ouroboros plans and owns technical authority for consequential work.
- Codex independently attacks and reviews proposals with evidence-backed challenges.
- Ouroboros rebuts, revises, or replaces challenged plans before Codex verifies again.

The supporting implementation artifacts and receipts are not included in this release.

## Not claimed

- No claim of consciousness, AGI, universal autonomy, or universal safety.
- No claim that every transition is completely mediated.
- This repository is not a public end-to-end implementation package.
- No quantitative claim that the architecture always outperforms a model-only agent.