OPS.16:5 - Archetypal Grounding

OPS.16:5.1 - PumpWorks branches an admitted package-admission Method

This constructed case continues the PumpWorks control-service operation. The incumbent package-admission Method is admitted PW-TestAdmission-v1. A source-traceable A.3.1.MR candidate-account episteme proposed adding current package readiness, required permission, qualified support, a downstream acceptance path, and a stop on new starts when incident margin is consumed. Before this trial, its EntityOfConcern was independently admitted as PW-TestAdmission-v2 : U.Method. Separate episteme PW-TestAdmission-Description-e2 : U.MethodDescription passed A.3.2 and is the exact description relied on here. The candidate account, admitted Method, and MethodDescription remain different objects.

The local hypothesis is that v2 will reduce avoidable setup and unsupported starts for the named control-service package family without damaging incident response or protected recovery. The alternatives are retain v1 for the whole family, use v2 for the named family, branch by an applicability condition, stop the change, or obtain further evidence. The control-service operations lead holds the local package-admission authority; safety and field-release decisions remain with their existing holders.

The prior OPS.19 result keeps the trial closed while incident coverage, recovery, support, permission or feasible rig use is absent. In the later trial interval, the service and resource owners confirm permitted overlap, retained incident coverage, manual fallback, a qualified support window, the stop on new starts and hand-back. Those conditions make a two-package trial feasible. The earlier plan could describe them while they were still unavailable; it could not authorize an unsupported start.

The two-package trial begins as a WorkPlan. It names two intended package-admission decisions, proposed coordinators, v2 as the admitted Method expected to be enacted, support and fallback, the service and recovery protections, the readiness and permission observations, the burden to record, and the stop. In the constructed later week, Coordinator-C17 performs PW-Admission-A-1 from 09:10 to 09:18 and Coordinator-C22 performs PW-Admission-B-1 from 10:05 to 10:12, both inside PumpWorks-ControlServiceOps. Each is separately admitted Work and enacts admitted v2.

The observations support the readiness and permission branch for these two decisions. They do not establish transfer to another service, causal superiority over v1, general reliability, or practical worth across a broader horizon. The operating decision can rely on the actual occurrences, available service protection and observed hold. If a later decision asks whether v2 causes fewer unsupported starts than v1, these observations do not supply the needed C.28 comparison and inference basis. Return that causal-support gap while preserving the independently supported branch decision.

The selected return is branch. Retain admitted PW-TestAdmission-v2 under PW-TestAdmission-Description-e2 for the named control-service package family and the supplied support and service conditions. Keep admitted v1 for an unaffected family whose current facts still fit. Before widening the v2 branch, require a discriminating observation under provider unavailability. Record the authorized operating decision, Method-lineage update, and repertoire return separately from the trial Work. Reopen when permission, provider support, incident margin, package family, relied-on description, or the next discriminating observation changes.

OPS.16:5.2 - A hospital keeps a proposed readiness way pre-admission

A public-hospital emergency-flow team observes repeated handover clarification and proposes a reusable readiness way for transferring a patient case between clinical areas. The provisionally distinguished way is carried by candidate account ED-Handover-Readiness-CA1. The account retains two serious explanations: missing readiness information may cause avoidable repeat work, while changing clinical priority or diagnostic uncertainty may explain the same delay. It names the missing identity evidence and the observation that could distinguish those accounts. No U.Method or U.MethodDescription has been admitted.

A bounded WorkPlan proposes observing two handover intervals under current staffing and safety conditions. It names the intended information check, qualified participants, privacy protection, stop, and the later identity decision. When actual clinical and coordination Work occurs, it is admitted separately. Only independently admitted triage, clinical-assessment, and handover Methods may be recorded as enacted. The candidate whole and its account are not enacted.

Suppose one repeated clarification follows missing readiness information while another follows a new clinical finding despite complete information. The first observation supports part of ED-Handover-Readiness-CA1; the second preserves the rival and shows that one fixed readiness rule cannot explain every delay. Patient, clinical case, WorkPlan, actual Work, information observations, safety conditions, and clinical decisions remain separate.

The current return is a revised candidate account with the rival and missing identity evidence retained. That revision is complete; it is not adopt or branch. If the later receiving decision needs Method identification, further discrimination may be needed before the separate A.3.1 test; select the inquiry only when its obtainable contribution warrants its whole burden. An unavailable observation leaves the identity claim unsupported without erasing the current account. Medical priority, consent, privacy, safety, staffing and worker-health decisions remain with their qualified owners.

OPS.16:5.3 - An AI-assisted software operation stops at a provider boundary

A software operations group has independently admitted AI-ReleasePrep-v3 : U.Method for one release-preparation family. Exact AI-ReleasePrep-Description-e4 : U.MethodDescription states the human review, evidence capture, security check, fallback, and release-authority boundary. A prompt repository, model endpoint, and usage counter are supporting Systems or records; none is the Method or proof of enactment.

The trial WorkPlan compares the current human preparation route with admitted v3 for a small set of eligible low-risk changes. It names actual human reviewers, the admitted Method expected to be enacted, model and provider condition, security and software-assurance checks, fallback, review burden, evidence window, and a stop if the provider or model condition changes. Actual release-preparation and review Work is admitted separately. Model calls, traces, generated text, test results, security observations, human decisions, and release authorization retain their own subjects.

For eligible cases under the qualified provider and model condition, the observations can support a bounded adopt or branch decision if human review, assurance, service, and burden conditions remain current. If the provider changes the model to an unqualified edition during the interval, the existing evidence no longer supports that branch. The team stops AI-assisted starts for the changed condition and uses the named fallback while qualification is unresolved.

The return preserves the admitted Method and its exact description while narrowing its applicable branch. It establishes neither model capability in general, causal superiority, release authority for the tool, nor transfer to high-risk changes. Security, software assurance, model capability, employment, and release decisions remain with their owners.