{"id":"ENG-3","type":"engagement","title":"Evaluation Stability Measurement (provisional)","created":"2026-07-13","status":"provisional","authors":["Upstream Zero"],"edges":[{"rel":"exercises","to":"CAP-1"}],"serves":"Organizations making decisions based on single-shot AI evaluation results","requiresFromClient":"The evaluation contexts that matter to you (categories, prompts, evaluator classes)","nonPromises":["We do not promise stability — we measure it, whatever it turns out to be"],"founderDecision":"FD-6","body":"How stable is what evaluators say about you? Repeated, paraphrased,\nmulti-condition sampling of evaluator behavior about your organization,\nreported with variance — because a recommendation that flips under\nre-phrasing is noise wearing a recommendation's clothes (see Q-3).","url":"/engagements/ENG-3","machineUrl":"/objects/ENG-3","referencedBy":[],"_meta":{"site":"Upstream Zero — Commercial Evaluation Observatory","version":"0.1","note":"Claims are presented at their evidence tier; Narrated is the lowest. Verify by walking edges, not by trusting us."}}