Private evidence plane · regression-aware workflows
An anonymous, session-bound workspace keeps your events private. Build a deterministic control cohort, then issue a portable integrity proof whose limits travel with the artifact.
Record deploy and configuration context next to provider observations. Mark an outcome explicitly; the analyzer never guesses degradation from prose. Nothing here is public research data.
No private events yet.
Only explicit degraded observations enter this view. Nearby deploys and configuration changes are labeled “coincided with,” with time and URL scope attached. The result is a review lead, not a causal finding.
Add a preflight, crawl, or index observation and mark its outcome degraded. Unknown and improved states remain outside the regression set.
72 hours · confidence capped at medium
SHA-256 assignment stays stable across reruns. Only treatment URLs receive the intervention; both cohorts receive the same observation method.
Canonical JSON, SHA-256 content digest, and ES256 signature. The generated key is ephemeral: verification proves integrity, not signer identity.
Consent is purpose-specific and revocable. The row contains only platform and size buckets, event/cohort labels, duration, and time—never a URL, domain, query, or provider payload. Public groups require five distinct workspaces.
