MagAI uses a physics-credible A / B / C / untrusted grading derived from five independent checks — mesh convergence, energy balance, reference benchmark, datasheet match, and lab measurement.
The grade is computed by the engine on every result, not assigned by hand. We publish enough of the framework to make the grade interpretable; the full methodology with thresholds, weights, and calibration data is shared on request with engineering reviewers, standards bodies, and pilot customers.
The complete 10-section methodology — component definitions, thresholds, reference benchmarks, and lab-calibration protocol — is available on request. Send a short note about your use case.