WINDLOADS.AI
Open Averroes
Validation disclosure

Public performance metrics are not yet available.

Windloads.ai has indexed more than 500 laboratory reports as contextual evidence. The archive has not yet been published as a frozen, independently reviewed evaluation set, so it does not by itself establish predictive accuracy or statistical calibration.

Current status

Forecasts remain model-based engineering judgment.

R², mean absolute error, false-positive rate, and false-negative rate will be released only after the evaluation set and inclusion rules are frozen. Until then, comparable laboratory cases provide context rather than a calibrated pass probability.

Publication rules

Every public metric must pass all six gates below.

  1. 01Freeze inclusion and exclusion rules before calculating metrics.
  2. 02Keep development specimens separate from an independent holdout set.
  3. 03Compare predicted and measured response at the same pressure and location.
  4. 04Publish sample size, R², mean absolute error, false-positive rate, and false-negative rate.
  5. 05Break results out by windows, doors, curtain wall, garage doors, and skylights.
  6. 06Disclose sparse or missing coverage for each system family.