First deliverable and evaluation boundary
Deliverable
A short account of how the current method allocates investigation and review effort, followed by options for managing the gap between available resources and possible source coverage. Address selection, depth, stopping, uncertainty and the cost of missed mechanisms. Compare the current role contracts with those needs and recommend the smallest sufficient change. Distinguish reducing avoidable work from deliberately leaving questions unanswered.
Start from current arrangement and the read-count baseline in evidence base. Compare at least these three options, none of which adds a process or touches the engine:
- Clarify the existing materiality-based stopping rule for omission searches, without changing coverage territory.
- Bound lens omission searches by runtime and lens anchors, as the verification coverage bounds proposal describes.
- Select a small, justified set of consequential claims for the existing report verifier's question-driven omission checks, with an explicit stopping decision and unresolved coverage. Do not invent a numerical budget.
The completed comparison recommends the third as an operational clarification of the first, not a new architecture. The operator adopted it with effort-sensitive stopping and brief reporting; the prose-only implementation is recorded in the comparison. The return condition below governed the initial comparison and still applies to any further trial.
The proposal is one option among these, not the conclusion.
Do not begin with a replacement workflow, a predetermined memory split or an assumption that more processes produce better coverage.
Evaluation criteria
- Whether the option reveals consequential design alternatives and mechanisms.
- Whether findings retain enough conditions and evidence to inform a later Commonplace transfer assessment.
- Whether the overall account remains coherent and claims stay within evidence.
- Total production, verification and consumption work, including repeated reads and cross-lens reconciliation.
- Disruption to the stabilized method, command environment, report consumers and existing tests.
Return condition
Return to the operator after the comparison, before any live method change or trial. A partial comparison names the unresolved choice and the evidence needed rather than silently selecting an architecture. Any proposed trial specifies its question, scope, cost, success criteria and the evidence that would count against adoption; running it requires separate authorization.