Public transparency

Progress you can inspect, limits you can see.

This page publishes the current methodology snapshot, reproducibility checks, and benchmark boundaries. It does not turn activity into claims about users, security, or outcomes.

Loading the current snapshot…

Current snapshot

What is established now

Published values are dated, versioned, and accompanied by their definition and limitation.

Interpretation. These are engineering and methodology indicators. A directive generated is not a person protected; a reported control is not a verified control; a benchmark result is not a real-world guarantee.

Reproducibility

A small benchmark, clearly bounded

The current benchmark tests the harness with synthetic fixture suggestions. No external model provider is enabled.

Adoption and review outcomes

Insufficient data is a valid result.

Aggregate reporting begins only when the published privacy threshold is met.

Reproduce it

Use the repository as the record.

The snapshot is backed by open schemas, fixtures, Control Packs, and reference validation. Run the same checks locally and compare methodology versions before comparing results.

python scripts/validate.pypython scripts/benchmark_semantic.pypython -m unittest discover -s tests -p "test_*.py"

Limitations and corrections

Evidence comes with boundaries.