Every public protocol carries four design-quality scores (feasibility, controls, reproducibility, and clarity) from its AI design review, each on a 0–100 scale. The “design score” shown here is the mean of those four.
Only protocols that carry a real (non-zero) score are included, so unscored drafts never drag the distribution down. Counts and averages are computed over exactly those public, scored protocols.
These are AI design-review scores. The deterministic, seven-signal reproducibility rigor score, which no AI touches and anyone can recompute, lives on each protocol’s Reproducibility Passport, not in the averages above.
Every figure is aggregated from the live open library and refreshes continuously (cached for a few minutes). Nothing here is a fabricated or seeded number: open any protocol in the library and verify its scores yourself.
Updated live · aggregated 2026-09-27. See how the deterministic score works →