Research Methodology

Source Materials

Content is derived from three source categories:

• Primary research: Published academic papers (see Research Index for 12 annotated sources)
• Architectural documentation: Carcinus.org source code, .uai memory system, deployment logs
• Design specifications: Internal framing documents, claim-status matrix, evaluation packet schema

All sources are cited or linked from the relevant pages. No content is generated by AI models — all claim-boundary text is human-authored and reviewed.

Claim-Status Assignment Process

Every public claim on Carcinus.org must carry one of 7 statuses from the Claim-Status Matrix. The assignment process follows these rules:

1. Framing claims are assigned when the claim defines scope or methodology — not an empirical result
2. Architectural claims are assigned when the claim describes a structural property of the platform
3. Implemented static support requires verifiable evidence that the capability exists in live infrastructure
4. Future handoff is assigned when the capability is specified but deliberately not implemented on this site
5. Research hypothesis covers speculative ideas that are worth testing but lack evidence
6. Not claimed and Rejected overclaim are defensive labels that explicitly bound what Carcinus.org does not assert

Versioning Strategy

All changes to the Teleodynamic AI section follow semantic versioning:

• Major (X.0.0): New research sections, fundamental claim changes, redesigned architecture
• Minor (0.X.0): New pages, new features, content expansion — each pass increments the minor version
• Patch (0.0.X): Corrections, accessibility fixes, broken link repairs, typo fixes

Version history is published in the Carcinus.org Changelog. Every commit is linked to a specific version number. See the Implementation Roadmap for planned future versions.

Peer Review Protocol

Before publication, each page is subject to a review checklist:

1. Verify all claims carry visible status labels consistent with the claim-status matrix
2. Confirm no prohibited hype language appears as a factual claim
3. Check all cross-links resolve correctly
4. Validate JSON-LD, OpenGraph, and SEO metadata
5. Review for accessibility: heading order, aria labels, contrast, keyboard navigation
6. Run automated validation tests (Node.js test suites) for prohibited phrases, JSON validity, and secret patterns
7. Human reviewer signs off through the Reviewer Toolkit process

Update Cadence & Correction Policy

Update cadence: Content updates follow a sweep-based model — each version pass delivers a complete, validated batch of changes. No incremental, unprompted edits occur between versions.

Correction policy: If an error in a claim or a broken link is discovered, it is corrected in the next minor version pass. Critical corrections (security, claim-boundary violations) are applied immediately as patches. All corrections are recorded in the changelog with the corrected version number.

Archival Strategy

Every version pass produces:

• A dated archive entry in .uai/archives/ — full session summary with pages, files, safety boundaries
• An updated .uai/progress.uai — cumulative progress record across all versions
• A production ZIP containing both the published build and source documentation
• A git commit with the version number and pass title

The Evidence Bundle Archive extends this concept to evaluation packet data, providing downloadable evidence bundles with checksums and manifests.

Explore Further