Skip to content

Clarethium

Methodology, standards, and reference implementations for AI-coupled work.

Clarethium develops model-independent verification methodology for AI outputs and the tools that implement it. The work is open: standards under CC-BY 4.0, libraries under Apache 2.0.


Projects

Model-independent verification for AI-coupled work. The Touchstone Standard defines eleven measurement layers for output profiling and five layers for specification compliance verification. The clarethium-touchstone library is the reference implementation.

Domain: touchstone.clarethium.com

Frame Check

Applied tool for frame validation in AI outputs. Uses Clarethium methodology in a specific applied context.

Domain: frame.clarethium.com


What "model-independent" means

Most AI-evaluation tools use AI models to evaluate AI output. This produces structural conflicts of interest: the auditor inherits the same biases and failure modes as the audited.

Clarethium methodology breaks the loop. The substrate uses regex, structural analysis, source comparison, and arithmetic. It runs without invoking AI models on the outputs being measured. AI does not obsolete arithmetic.

This matters because:

  • Trust requires verifiability
  • Independence is structural, not an aspiration
  • Time-locked deterministic measurement compounds where AI-judged measurement does not

Methodology principles

The work rests on four commitments:

  1. Model-independence. Required measurements operate without AI on the output.
  2. Structural over semantic. Reproducible structural properties over judgment about meaning.
  3. Calibrated thresholds. Explicit, versioned, overridable defaults.
  4. Open and verifiable. Standards are CC-BY; libraries are Apache 2.0; reference test cases are public.

How to engage

  • Read the Touchstone Standard if you want to understand the methodology.
  • Use the libraries to verify AI outputs in your work.
  • Cite the Standard when you use the methodology in research or compliance.
  • Propose changes via the Suggestion Process.
  • Build conforming implementations in other languages or for specific use cases.

Related

  • FieldReceipts (fieldreceipts.com) — community platform for AI-coupled builders, uses Clarethium methodology as substrate. Independent project.
  • Clarethium website (clarethium.com) — methodology research, blog, mothership.

Status

Pre-launch for the Touchstone Standard 1.0 (drafting in progress). Frame Check operational. Other projects in development.

Watch the Touchstone repository for first release.


The work is open. The substrate is verifiable. The institution stays independent.

Popular repositories Loading

  1. touchstone touchstone Public

    Python 1

  2. .github .github Public

Repositories

Showing 2 of 2 repositories

People

This organization has no public members. You must be a member to see who’s a part of this organization.

Top languages

Loading…

Most used topics

Loading…