build1 distinct publisher
Tripling the frame count flipped a pre-registered pass/fail verdict on the same clip
The bar was 50% of the original motion median, and the same clip honestly returned 47.5% and 56.5%, because the registration fixed the bar and left the sample count to the pipeline. The rule written to close that gap then produced a compliance rate with the same hole.
Publishers:dev.to
Reality
- Evidence58
- Adoption16
- Hype gap−9
- Incentives31