Case studies with the uncertainty left in.
Each note is one real engagement: the prompt that was missing, the fix that shipped, and dated re-test observations of the same prompt afterward. The Proof Ledger publishes the aggregates; these are the individual stories - wins, early signals, and no-movement results alike.
Collecting
The first notes are in flight: RankEcho is being run on our own properties and a small set of consenting external sites, fixes are shipping, and the re-test clocks are running. Notes publish here as observations accumulate - including the ones where nothing moved.
Until then: the proof loop demo shows the format on deterministic sample data, and the Proof Ledger explains the aggregate methodology.
What every note contains
One engagement, told end to end, in the same order every time:
- The missing prompt and the baseline engine result - including who was cited instead.
- The fix that shipped, dated, listing exactly what was deployed.
- Dated re-test observations - sampled runs of the same prompt on the same engines.
- The movement summary - and "no sampled movement yet" gets published too.
- An explicit uncertainty note. Engine answers are stochastic; a note without stated uncertainty does not ship.
Why publish the misses? Because a proof system that only shows wins is a testimonial page. The point of the proof loop is that the before/after lives in dated observations either way - that is what makes the wins believable.