Method proposal · Zero live contributions · No dataset is being collected
Agent Service ProviderCompatibility Lab

Publish patterns only after people consent and the sample is large enough.

The proposed dataset will count coarse compatibility blocker categories. It will never publish submitted manifests, service names, endpoints, report digests, contact details, or raw examples.

Current dataset status: not started. This public beta has no analytics, research submission, or cross-user data store.

Allowed research record

One consented contribution may contain only the report family, coarse protocol family, compatible or blocked state, the canonical blocker dimensions, and a broad acquisition channel. RunSAVE structural signals remain a separate report family and are never mixed with Nexus compatibility blockers.

Never collected for research

Manifest or graph JSON, names, labels, URLs, filenames, free text, validation messages, exact referrers, timestamps, user-agent strings, IP addresses, fingerprints, or any digest.

Publication threshold

  • At least 100 explicitly consented contributions across at least ten organisations.
  • At most one current contribution per confirmed participant, report family, and quarter.
  • Suppress every cell with fewer than 30 contributions or ten organisations.
  • Delay publication by at least 30 days and round counts and rates.
  • Keep any one organisation below 20% of the published sample.
  • Describe the population as a self-selected pilot sample, never an industry benchmark.

Joining a provider preview and contributing diagnostic categories are different choices. A future form must leave research consent unchecked, show the exact coarse record before submission, provide a withdrawal route, and define retention and deletion.

No page will claim “common provider problems” until this threshold is met and the aggregate has received privacy and human content review.

Method proposal recorded . It is not a published Nexus machine contract.