Manifold Grace

WORKING ARCHITECTURE · NOVEMBER 2026 · PROPOSED INTEGRATIONS ARE NOT DEPLOYMENTS

Modular AI Platform: Projects, Components and Connections

Existing research, demonstrations and project specifications can become interoperable components. This map preserves independent projects while identifying shared capabilities and candidate technical connections. Project identity remains stable if branding changes.

Original project register · Existing implementation roadmap · Machine-readable project register · Architecture method

1. Platform layers

Knowledge: archives, research discovery and question records. Evidence: validation, dated claims and provenance. Intelligence: model access and analysis. Assurance: testing, review and remediation. Orchestration: authorised workflows. Applications: business, education, societal services and accessible interfaces.

These are conceptual groupings, not running integrated services.

2. Selected existing project records

IDExisting projectFamilyEvidence status

3. Proposed project-to-project connections

All links below are architectural suggestions. No API wiring or tested data flow is implied.

FromToProposed interface purposeStatus

4. Missing shared capabilities to evaluate

These should be created only after a source-code inventory confirms they do not already exist.

5. First integration candidate

Research discovery → Evidence verification → Continuous AI assurance → Accessible results presentation. Required gates: source-code inventory, documented schemas, privacy review, contract tests, regression tests, reviewer sign-off and deployment verification.

Evidence hierarchy: documented page ≠ source-code implementation ≠ tested integration ≠ live service. Keep each independently verifiable.

6. Change and provenance record

ItemEvidence linkWhy it mattersNext step
14 November 2026 — Initial modular architecture registerVersioned project JSONPreserves existing work and separates proposed links from verified connectionsInspect source and test first integration candidate

Product development — AI Assurance Studio

The QA research methodology is being developed into a separate product with an account gateway, private project workspace, and source-linked QA test catalogue. Explore the product. Registration requires backend deployment and verification; it is not active yet.